diff --git a/dsgai_scanner_tool/CHANGES_v0.3.md b/dsgai_scanner_tool/CHANGES_v0.3.md
index 86ab567..85efbeb 100644
--- a/dsgai_scanner_tool/CHANGES_v0.3.md
+++ b/dsgai_scanner_tool/CHANGES_v0.3.md
@@ -17,6 +17,13 @@ dates are ISO-8601. The previous line is recorded in [`CHANGES_v0.2.md`](CHANGES
   internal-link check), a deterministic markdown internal-link checker
   (`scripts/check_md_links.py`), and a `.gitattributes` forcing LF on scripts/YAML so
   Windows checkouts can't ship CRLF that breaks Linux CI. (PR-02)
+- **Rules as data**: all 106 detection patterns extracted verbatim from the skill's
+  Step 2 into `rules/dsgai-rules.yaml` (source of truth), validated by
+  `rules/rules.schema.json`, compiled to `rules/dsgai-rules.json` by
+  `build/build_rules_json.py`. Compound logic (`subtract`, `requires_nearby`,
+  `exclude_globs`, `gated_on`) and per-rule `classification`/`signal`/`confidence` are
+  encoded. `rules/README.md` documents the format. Skill Step 2 now marks the YAML as
+  canonical. (PR-03)
 
 ### Changed
 - `DSGAI-samplereport.png` compressed from ~5.0 MB to ~0.35 MB (14×) as an interim fix;
diff --git a/dsgai_scanner_tool/CONTRIBUTING.md b/dsgai_scanner_tool/CONTRIBUTING.md
index 09615f8..c1951c2 100644
--- a/dsgai_scanner_tool/CONTRIBUTING.md
+++ b/dsgai_scanner_tool/CONTRIBUTING.md
@@ -28,11 +28,11 @@ make the tool measurably better — a good repro is the contribution.
 
 ## Contributing a rule
 
-Detection rules are moving from prose in `dsgai_scanner_tool.md` into data at
-`rules/dsgai-rules.yaml`, validated by `rules/rules.schema.json`. **(Landing soon —
-tracked by PR-03.)** Until that lands, describe your rule in a
-[new-rule issue](../.github/ISSUE_TEMPLATE/scanner-new-rule.yml) using the format below;
-once the YAML rule format ships, this section will point at `rules/README.md`.
+Detection rules live as data in [`rules/dsgai-rules.yaml`](rules/dsgai-rules.yaml),
+validated by `rules/rules.schema.json` and compiled to `rules/dsgai-rules.json`. See
+[`rules/README.md`](rules/README.md) for the full field reference. You can also propose
+a rule without writing YAML via a
+[new-rule issue](../.github/ISSUE_TEMPLATE/scanner-new-rule.yml).
 
 Every rule needs:
 
diff --git a/dsgai_scanner_tool/build/build_rules_json.py b/dsgai_scanner_tool/build/build_rules_json.py
new file mode 100644
index 0000000..2ac6fcc
--- /dev/null
+++ b/dsgai_scanner_tool/build/build_rules_json.py
@@ -0,0 +1,56 @@
+#!/usr/bin/env python3
+"""Build rules/dsgai-rules.json from rules/dsgai-rules.yaml.
+
+This is the repeatable build step (unlike the one-time generate_rules.py). The
+deterministic CLI loads the JSON with the standard library only, so the
+curl-one-file install story needs no PyYAML at runtime. PyYAML is required only
+to edit rules and to run this build / the self-test that asserts JSON == YAML.
+
+Validates against rules/rules.schema.json before writing. Output is
+deterministic (sorted keys, stable rule order, trailing newline) so the PR-06
+self-test can assert the checked-in JSON matches a fresh build.
+
+Usage:
+  python build/build_rules_json.py            # write rules/dsgai-rules.json
+  python build/build_rules_json.py --check     # verify checked-in JSON is current (CI)
+"""
+import json
+import sys
+from pathlib import Path
+
+RULES_DIR = Path(__file__).resolve().parent.parent / "rules"
+YAML_PATH = RULES_DIR / "dsgai-rules.yaml"
+JSON_PATH = RULES_DIR / "dsgai-rules.json"
+SCHEMA_PATH = RULES_DIR / "rules.schema.json"
+
+
+def build():
+    import yaml  # only needed for build/edit, not at CLI runtime
+    data = yaml.safe_load(YAML_PATH.read_text(encoding="utf-8"))
+    try:
+        import jsonschema
+        schema = json.loads(SCHEMA_PATH.read_text(encoding="utf-8"))
+        jsonschema.validate(data, schema)
+    except ImportError:
+        sys.stderr.write("warning: jsonschema not installed; skipping validation\n")
+    return json.dumps(data, indent=2, sort_keys=True, ensure_ascii=False) + "\n"
+
+
+def main(argv):
+    rendered = build()
+    if "--check" in argv:
+        current = JSON_PATH.read_text(encoding="utf-8") if JSON_PATH.exists() else ""
+        if current != rendered:
+            sys.stderr.write(
+                "rules/dsgai-rules.json is out of date. Run: "
+                "python build/build_rules_json.py\n")
+            return 1
+        print("rules/dsgai-rules.json is up to date.")
+        return 0
+    JSON_PATH.write_text(rendered, encoding="utf-8", newline="\n")
+    print(f"Wrote {JSON_PATH} ({rendered.count(chr(10))} lines).")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main(sys.argv[1:]))
diff --git a/dsgai_scanner_tool/build/generate_rules.py b/dsgai_scanner_tool/build/generate_rules.py
new file mode 100644
index 0000000..36b5808
--- /dev/null
+++ b/dsgai_scanner_tool/build/generate_rules.py
@@ -0,0 +1,233 @@
+#!/usr/bin/env python3
+"""One-time bootstrap: extract DSGAI scanner patterns from dsgai_scanner_tool.md
+Step 2 into rules/dsgai-rules.yaml.
+
+Patterns are copied VERBATIM (bug-for-bug faithful — fixes come in PR-11 as
+reviewable diffs). Classification, signal, confidence, and compound logic
+(subtract / requires_nearby / exclude_globs / gated_on / notes) are augmented
+from the skill's prose and the improvement plan's Appendix B seeds.
+
+After this bootstrap runs, rules/dsgai-rules.yaml is the hand-maintained source
+of truth; the skill prose becomes descriptive. Re-running would overwrite manual
+edits, so it is kept only for provenance / audit.
+
+Usage: python build/generate_rules.py > rules/dsgai-rules.yaml
+"""
+import re
+import sys
+from pathlib import Path
+
+SKILL = Path(__file__).resolve().parent.parent / "dsgai_scanner_tool.md"
+VALUE_BEARING_CONTROLS = {2, 13, 14, 15}
+FRAMEWORK = "dsgai-2026-v1.0"
+RULESET_VERSION = "0.3.0"
+
+HEADER_RE = re.compile(r'^### DSGAI(\d{2}) Scan .*\[(STRUCTURAL|VALUE-BEARING)')
+FILES_RE = re.compile(r'^Files:\s*(.+)$')
+# Pattern line: ID  <desc up to first colon>: <pcre to EOL>.  ':' separator is
+# the FIRST colon — no rule description contains an internal colon.
+PAT_RE = re.compile(r'^(P\d{2}\.\d+)\s+(.+?):\s+(.+?)\s*$')
+GLOB_RE = re.compile(r'`([^`]+)`')
+
+# Signal overrides for rules whose FAIL/PASS status comes from prose, not a
+# marker in the description text.
+SIGNAL_OVERRIDE = {
+    "P02.1": "fail", "P02.2": "fail", "P02.3": "fail",
+    "P02.4": "fail", "P02.5": "fail",
+}
+
+# Compound logic and gating, transcribed from the Step 2 prose notes.
+SUBTRACT = {"P04.1": ["P04.2"]}
+REQUIRES_NEARBY = {
+    "P05.1": {"rules": ["P05.2", "P05.3"], "scope": "module"},
+    "P06.5": {"rule": "P06.2", "scope": "module", "absent": True},
+    "P11.1": {"rule": "P11.2", "lines": 15},
+    "P18.4": {"rule": "P18.5", "lines": 10},
+    "P20.5": {"rules": ["P20.1", "P20.2"], "lines": 15},
+}
+EXCLUDE_GLOBS = {
+    "P12.6": ["**/migrations/**", "**/fixtures/**", "**/tests/**", "**/test/**"],
+}
+GATED_ON = {
+    "P09.1": "multimodal", "P09.2": "multimodal", "P09.3": "multimodal",
+    "P09.4": "multimodal", "P09.5": "multimodal",
+    "P10.1": "synthetic_data", "P10.2": "synthetic_data", "P10.3": "synthetic_data",
+    "P10.4": "synthetic_data", "P10.5": "synthetic_data",
+    "P19.1": "labeling", "P19.2": "labeling", "P19.3": "labeling", "P19.4": "labeling",
+}
+# Free-text prose that resists full formalization at import time (refined later).
+NOTES = {
+    "P07.1": "Absence of P07.1-P07.4 in a multi-tenant or PII-handling repo = WARN.",
+    "P08.1": "Absence of P08.1-P08.6 in a production GenAI service = WARN; "
+             "absence in a high-risk EU AI Act use case = FAIL.",
+    "P09.5": "Absence = note only (advanced control); P09.1-P09.4 absence in a "
+             "multimodal pipeline = WARN.",
+    "P10.1": "Synthetic data pipeline without any of P10.1/P10.2/P10.4 = FAIL.",
+    "P12.6": "DDL in migrations/fixtures/tests is benign — excluded from FAIL.",
+    "P14.4": "May contain inline PII in the format string — treated as VALUE-BEARING.",
+    "P16.1": "Filename-existence check (not content grep). Absence in a repo with "
+             ".env or secrets/ = WARN.",
+    "P16.2": "Matched inside any AI-ignore file.",
+    "P17.1": "LLM-calling module with none of P17.1-P17.5 = WARN.",
+    "P21.2": "P21.2 in an agent module without P21.3 = WARN.",
+}
+# P16 rules carry a mode prefix ('filename match:' / 'in any ignore file:') in
+# the pcre slot; strip it to the bare regex.
+MODE_PREFIX_RE = re.compile(r'^(filename match|in any ignore file):\s*')
+
+
+def slugify(text):
+    text = re.sub(r'\([^)]*\)', '', text)          # drop (FAIL)/(PASS)/... markers
+    text = text.lower().strip()
+    text = re.sub(r'[^a-z0-9]+', '-', text).strip('-')
+    return text or "rule"
+
+
+def derive_signal(pid, desc):
+    if pid in SIGNAL_OVERRIDE:
+        return SIGNAL_OVERRIDE[pid]
+    d = desc.lower()
+    if "fail" in d:
+        return "fail"
+    if "warn" in d:
+        return "warn"
+    if "pass" in d:
+        return "pass_signal"
+    if "count" in d:
+        return "count"
+    return "info"
+
+
+def derive_confidence(pid, control, signal, pcre):
+    # Value-bearing credential-literal FAILs are the high-confidence anchors.
+    if control in VALUE_BEARING_CONTROLS and signal == "fail":
+        return "high"
+    if signal == "warn":
+        return "low"          # heuristic / absence-adjacent — weak evidence
+    if signal in ("pass_signal", "count"):
+        return "medium"       # an import is not proof of correct use
+    if signal == "fail":
+        return "medium"       # structural heuristic FAILs (e.g. P12.1, P17.6)
+    return "low"              # bare detection/info
+
+
+def yaml_scalar(s):
+    """Single-quote a scalar for YAML, escaping embedded single quotes."""
+    return "'" + s.replace("'", "''") + "'"
+
+
+def emit_list(vals):
+    return "[" + ", ".join(yaml_scalar(v) for v in vals) + "]"
+
+
+def main():
+    # Ensure UTF-8 output regardless of the platform console codepage (Windows
+    # defaults to cp1252, which corrupts em-dashes / non-ASCII on redirect).
+    try:
+        sys.stdout.reconfigure(encoding="utf-8", newline="\n")
+    except AttributeError:
+        pass
+    lines = SKILL.read_text(encoding="utf-8").splitlines()
+    control = None
+    classification = None
+    file_globs = []
+    in_fence = False
+    rules = []
+    unparsed = []
+
+    # Only scan Step 2 (between its header and Step 3).
+    start = next(i for i, l in enumerate(lines) if l.startswith("## Step 2:"))
+    end = next(i for i, l in enumerate(lines) if l.startswith("## Step 3:"))
+
+    for raw in lines[start:end]:
+        line = raw.rstrip("\n")
+        m = HEADER_RE.match(line)
+        if m:
+            control = int(m.group(1))
+            classification = "value_bearing" if control in VALUE_BEARING_CONTROLS else "structural"
+            file_globs = []
+            in_fence = False
+            continue
+        fm = FILES_RE.match(line)
+        if fm:
+            file_globs = GLOB_RE.findall(fm.group(1))
+            continue
+        if line.strip().startswith("```"):
+            in_fence = not in_fence
+            continue
+        if in_fence and control is not None:
+            pm = PAT_RE.match(line)
+            if not pm:
+                if line.strip():
+                    unparsed.append(line)
+                continue
+            pid, desc, pcre = pm.group(1), pm.group(2).strip(), pm.group(3)
+            pcre = MODE_PREFIX_RE.sub("", pcre).strip()
+            signal = derive_signal(pid, desc)
+            rules.append({
+                "id": pid,
+                "control": f"DSGAI{control:02d}",
+                "name": slugify(desc),
+                "classification": classification,
+                "signal": signal,
+                "confidence": derive_confidence(pid, control, signal, pcre),
+                "pcre": pcre,
+                "file_globs": list(file_globs),
+                "exclude_globs": EXCLUDE_GLOBS.get(pid, []),
+                "framework": FRAMEWORK,
+                "description": desc,
+                "subtract": SUBTRACT.get(pid),
+                "requires_nearby": REQUIRES_NEARBY.get(pid),
+                "gated_on": GATED_ON.get(pid),
+                "notes": NOTES.get(pid),
+            })
+
+    if unparsed:
+        sys.stderr.write("UNPARSED LINES:\n" + "\n".join(unparsed) + "\n")
+
+    # Emit YAML by hand (deterministic ordering, faithful quoting of PCREs).
+    out = []
+    out.append("# DSGAI scanner detection rules — source of truth.")
+    out.append("# Generated once from dsgai_scanner_tool.md Step 2 by build/generate_rules.py,")
+    out.append("# then hand-maintained. Validated by rules/rules.schema.json (see rules/README.md).")
+    out.append(f"ruleset_version: '{RULESET_VERSION}'")
+    out.append(f"framework: '{FRAMEWORK}'")
+    out.append("rules:")
+    for r in rules:
+        out.append(f"  - id: {r['id']}")
+        out.append(f"    control: {r['control']}")
+        out.append(f"    name: {r['name']}")
+        out.append(f"    classification: {r['classification']}")
+        out.append(f"    signal: {r['signal']}")
+        out.append(f"    confidence: {r['confidence']}")
+        out.append(f"    pcre: {yaml_scalar(r['pcre'])}")
+        out.append(f"    file_globs: {emit_list(r['file_globs'])}")
+        out.append(f"    exclude_globs: {emit_list(r['exclude_globs'])}")
+        out.append(f"    framework: '{r['framework']}'")
+        out.append(f"    description: {yaml_scalar(r['description'])}")
+        if r["subtract"]:
+            out.append(f"    subtract: {emit_list(r['subtract'])}")
+        if r["requires_nearby"]:
+            rn = r["requires_nearby"]
+            parts = []
+            if "rule" in rn:
+                parts.append(f"rule: {rn['rule']}")
+            if "rules" in rn:
+                parts.append("rules: " + emit_list(rn["rules"]))
+            if "lines" in rn:
+                parts.append(f"lines: {rn['lines']}")
+            if "scope" in rn:
+                parts.append(f"scope: {rn['scope']}")
+            if rn.get("absent"):
+                parts.append("absent: true")
+            out.append("    requires_nearby: {" + ", ".join(parts) + "}")
+        if r["gated_on"]:
+            out.append(f"    gated_on: {r['gated_on']}")
+        if r["notes"]:
+            out.append(f"    notes: {yaml_scalar(r['notes'])}")
+    sys.stdout.write("\n".join(out) + "\n")
+    sys.stderr.write(f"\nExtracted {len(rules)} rules.\n")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/dsgai_scanner_tool/dsgai_scanner_tool.md b/dsgai_scanner_tool/dsgai_scanner_tool.md
index 12bf0d8..b343a4e 100644
--- a/dsgai_scanner_tool/dsgai_scanner_tool.md
+++ b/dsgai_scanner_tool/dsgai_scanner_tool.md
@@ -564,6 +564,8 @@ Also identify:
 
 ## Step 2: Scan for DSGAI Issues
 
+> **Canonical rule definitions live in [`rules/dsgai-rules.yaml`](rules/dsgai-rules.yaml)** (validated by `rules/rules.schema.json`, compiled to `rules/dsgai-rules.json`). The pattern listings in this Step are descriptive — the YAML is authoritative and is what the deterministic CLI executes. When they disagree, the YAML wins. (Full skill rewrite to CLI-first orchestration is PR-07.)
+
 ### Search Engine Prerequisite
 
 All patterns below use **PCRE / Perl-compatible regex syntax** — `\s`, `{n,m}`, character classes inside groups, alternation. Inside Claude Code, the Grep tool (ripgrep) supports this natively. Outside Claude Code, use `rg` (ripgrep) or `grep -P` (GNU grep with PCRE). Plain POSIX BRE/ERE will *not* match `\s`, `\d`, or `{n,m}` correctly and will produce false negatives.
diff --git a/dsgai_scanner_tool/rules/README.md b/dsgai_scanner_tool/rules/README.md
new file mode 100644
index 0000000..7e9ba49
--- /dev/null
+++ b/dsgai_scanner_tool/rules/README.md
@@ -0,0 +1,61 @@
+# DSGAI scanner rules
+
+`dsgai-rules.yaml` is the **source of truth** for every detection pattern. It is
+validated by `rules.schema.json` and compiled to `dsgai-rules.json` (which the
+deterministic CLI loads with the standard library only — no PyYAML at runtime).
+
+- **Edit** `dsgai-rules.yaml`, then rebuild the JSON:
+  `python build/build_rules_json.py`
+- **Never edit** `dsgai-rules.json` by hand — it is generated. CI checks it is
+  in sync (`python build/build_rules_json.py --check`).
+
+## Rule schema
+
+```yaml
+- id: P02.1                      # P<control>.<n>, unique
+  control: DSGAI02               # DSGAI01 .. DSGAI21
+  name: hardcoded-openai-api-key # kebab-case slug
+  classification: value_bearing  # structural | value_bearing
+  signal: fail                   # fail | warn | pass_signal | count | info
+  confidence: high               # high | medium | low
+  pcre: '(?i)(OPENAI_API_KEY|...)\s*[:=]\s*["'']?sk-[A-Za-z0-9_\-]{20,}'
+  file_globs: ['*.py', '*.env*'] # which files the rule runs against
+  exclude_globs: []              # paths excluded from this rule
+  framework: dsgai-2026-v1.0     # framework version binding
+  description: 'Hardcoded OpenAI API key assignment'
+```
+
+### Field reference
+
+| Field | Required | Meaning |
+|---|---|---|
+| `id` | yes | `P<NN>.<n>`, matches the control number |
+| `control` | yes | `DSGAI01`–`DSGAI21` |
+| `name` | yes | kebab-case identifier |
+| `classification` | yes | `structural` (match may be shown) or `value_bearing` (match content is a secret/PII — never shown; located in `--replace ''` mode) |
+| `signal` | yes | how a hit is weighted: `fail`, `warn`, `pass_signal`, `count`, `info` |
+| `confidence` | yes | `high` / `medium` / `low` — feeds SARIF `level` and report rendering |
+| `pcre` | yes | PCRE2 pattern; must compile under `rg --pcre2` (checked in CI, not by the schema) |
+| `file_globs` | yes | globs the rule scans |
+| `exclude_globs` | no | globs excluded from the rule (e.g. migrations/tests) |
+| `framework` | yes | `dsgai-YYYY-vX.Y` binding |
+| `description` | yes | human-readable summary |
+| `remediation` | no | fix guidance |
+| `references` | no | CVE IDs / links |
+| `subtract` | no | rule IDs whose match on the same line cancels this hit (e.g. `torch.load(` minus `weights_only=True`) |
+| `requires_nearby` | no | compound proximity logic: `{rule\|rules, lines\|scope, absent}` |
+| `gated_on` | no | only evaluated when the stack is detected: `multimodal`, `synthetic_data`, `labeling` |
+| `notes` | no | control-level prose that isn't yet fully formalized |
+
+## Classification: STRUCTURAL vs VALUE-BEARING
+
+See [`../CONTRIBUTING.md`](../CONTRIBUTING.md#structural-vs-value-bearing). In
+short: if a rule can match a line whose *content is* a secret or PII, it is
+`value_bearing` and runs in location-only mode so the value never leaves
+ripgrep. All rules under DSGAI02/13/14/15 are value-bearing.
+
+## Provenance
+
+The initial ruleset was extracted verbatim from `dsgai_scanner_tool.md` Step 2 by
+`build/generate_rules.py` (a one-time bootstrap, kept for audit). Pattern *fixes*
+land as reviewable diffs against this baseline starting in PR-11.
diff --git a/dsgai_scanner_tool/rules/dsgai-rules.json b/dsgai_scanner_tool/rules/dsgai-rules.json
new file mode 100644
index 0000000..d5dcc95
--- /dev/null
+++ b/dsgai_scanner_tool/rules/dsgai-rules.json
@@ -0,0 +1,2218 @@
+{
+  "framework": "dsgai-2026-v1.0",
+  "rules": [
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI01",
+      "description": "PII scrubbing imports",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P01.1",
+      "name": "pii-scrubbing-imports",
+      "pcre": "(anonymize|redact|scrub_pii|piiDetect|mask_pii|presidio)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI01",
+      "description": "No-train flags",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P01.2",
+      "name": "no-train-flags",
+      "pcre": "(allow_training|training_opt_out|X-Training-Data|no_train)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI01",
+      "description": "Output PII filter",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P01.3",
+      "name": "output-pii-filter",
+      "pcre": "(filter_pii|remove_pii|output_sanitize|output_filter)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI01",
+      "description": "Differential privacy lib",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P01.4",
+      "name": "differential-privacy-lib",
+      "pcre": "(opacus|tensorflow\\.privacy|dp-accounting|differential\\.privacy)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI01",
+      "description": "Consent / GDPR check",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P01.5",
+      "name": "consent-gdpr-check",
+      "pcre": "(consent_capture|gdpr|ccpa|data_subject|lawful_basis)",
+      "signal": "info"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "high",
+      "control": "DSGAI02",
+      "description": "Hardcoded LLM API key (OpenAI)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.js",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.env*",
+        "*.yaml",
+        "*.yml",
+        "*.json",
+        "*.toml",
+        "*.cfg"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P02.1",
+      "name": "hardcoded-llm-api-key",
+      "pcre": "(?i)(OPENAI_API_KEY|openai[._-]?api[._-]?key)\\s*[:=]\\s*[\"']sk-[A-Za-z0-9_\\-]{20,}",
+      "signal": "fail"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "high",
+      "control": "DSGAI02",
+      "description": "Hardcoded LLM API key (Anthropic)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.js",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.env*",
+        "*.yaml",
+        "*.yml",
+        "*.json",
+        "*.toml",
+        "*.cfg"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P02.2",
+      "name": "hardcoded-llm-api-key",
+      "pcre": "(?i)(ANTHROPIC_API_KEY|anthropic[._-]?api[._-]?key)\\s*[:=]\\s*[\"']sk-ant-[A-Za-z0-9_\\-]{20,}",
+      "signal": "fail"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "high",
+      "control": "DSGAI02",
+      "description": "Hardcoded Cohere/Google/HF tokens",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.js",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.env*",
+        "*.yaml",
+        "*.yml",
+        "*.json",
+        "*.toml",
+        "*.cfg"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P02.3",
+      "name": "hardcoded-cohere-google-hf-tokens",
+      "pcre": "(?i)(COHERE_API_KEY|GOOGLE_API_KEY|HF_TOKEN|HUGGINGFACE_TOKEN)\\s*[:=]\\s*[\"'][A-Za-z0-9_\\-]{20,}",
+      "signal": "fail"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "high",
+      "control": "DSGAI02",
+      "description": "Hardcoded AWS creds",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.js",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.env*",
+        "*.yaml",
+        "*.yml",
+        "*.json",
+        "*.toml",
+        "*.cfg"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P02.4",
+      "name": "hardcoded-aws-creds",
+      "pcre": "(AWS_ACCESS_KEY_ID|aws_access_key_id)\\s*[:=]\\s*[\"'][A-Z0-9]{16,}",
+      "signal": "fail"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "high",
+      "control": "DSGAI02",
+      "description": "Hardcoded Azure/GCP cred",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.js",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.env*",
+        "*.yaml",
+        "*.yml",
+        "*.json",
+        "*.toml",
+        "*.cfg"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P02.5",
+      "name": "hardcoded-azure-gcp-cred",
+      "pcre": "(AZURE_OPENAI_KEY|GCP_SERVICE_ACCOUNT_KEY)\\s*[:=]\\s*[\"'][A-Za-z0-9_\\-]{16,}",
+      "signal": "fail"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "low",
+      "control": "DSGAI02",
+      "description": "Wildcard token scope (warn)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.js",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.env*",
+        "*.yaml",
+        "*.yml",
+        "*.json",
+        "*.toml",
+        "*.cfg"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P02.6",
+      "name": "wildcard-token-scope",
+      "pcre": "(\"scope\"\\s*:\\s*\"\\*|permissions[^\\n]{0,30}\\*|scope[^\\n]{0,20}admin)",
+      "signal": "warn"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "medium",
+      "control": "DSGAI02",
+      "description": "Vault/secrets-manager (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.js",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.env*",
+        "*.yaml",
+        "*.yml",
+        "*.json",
+        "*.toml",
+        "*.cfg"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P02.7",
+      "name": "vault-secrets-manager",
+      "pcre": "(hvac\\.Client|hashicorp/vault|VAULT_(ADDR|TOKEN|NAMESPACE)|secretsmanager\\.|GetSecretValue|SecretManagerServiceClient|azure[._-]keyvault|@aws-sdk/client-secrets-manager)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "medium",
+      "control": "DSGAI02",
+      "description": "Tool-call signing (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.js",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.env*",
+        "*.yaml",
+        "*.yml",
+        "*.json",
+        "*.toml",
+        "*.cfg"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P02.8",
+      "name": "tool-call-signing",
+      "pcre": "(hmac|sign_request|verify_signature|tool_auth|mtls|mutual_tls)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI03",
+      "description": "Third-party LLM endpoints",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P03.1",
+      "name": "third-party-llm-endpoints",
+      "pcre": "(api\\.openai\\.com|api\\.anthropic\\.com|generativelanguage\\.googleapis\\.com|api\\.cohere\\.com|api\\.together\\.xyz|api\\.mistral\\.ai|api\\.groq\\.com)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI03",
+      "description": "Internal LLM gateway (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P03.2",
+      "name": "internal-llm-gateway",
+      "pcre": "(llm[._-]?gateway|ai[._-]?proxy|model[._-]?gateway|llm[._-]?proxy)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI03",
+      "description": "DLP / classification check",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P03.3",
+      "name": "dlp-classification-check",
+      "pcre": "(dlp|data_classification|classify_data|sensitivity_check)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI04",
+      "description": "Unsafe pickle (FAIL)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.sh",
+        "*.yml",
+        "*.yaml",
+        "Dockerfile",
+        "requirements*.txt",
+        "pyproject.toml",
+        "setup.py"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P04.1",
+      "name": "unsafe-pickle",
+      "pcre": "torch\\.load\\s*\\(",
+      "signal": "fail",
+      "subtract": [
+        "P04.2"
+      ]
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI04",
+      "description": "Safe pickle (PASS counter)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.sh",
+        "*.yml",
+        "*.yaml",
+        "Dockerfile",
+        "requirements*.txt",
+        "pyproject.toml",
+        "setup.py"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P04.2",
+      "name": "safe-pickle",
+      "pcre": "torch\\.load\\s*\\([^)]*weights_only\\s*=\\s*True",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI04",
+      "description": "Artifact verification (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.sh",
+        "*.yml",
+        "*.yaml",
+        "Dockerfile",
+        "requirements*.txt",
+        "pyproject.toml",
+        "setup.py"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P04.3",
+      "name": "artifact-verification",
+      "pcre": "(sha256|verify_signature|model_hash|check_integrity)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI04",
+      "description": "Unpinned ML deps (WARN)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.sh",
+        "*.yml",
+        "*.yaml",
+        "Dockerfile",
+        "requirements*.txt",
+        "pyproject.toml",
+        "setup.py"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P04.4",
+      "name": "unpinned-ml-deps",
+      "pcre": "^(torch|transformers|tensorflow|langchain|openai|anthropic|llama-index)\\s*(>=|~=|\\^|>|<|<=|\\*|latest)",
+      "signal": "warn"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI04",
+      "description": "SBOM in CI (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.sh",
+        "*.yml",
+        "*.yaml",
+        "Dockerfile",
+        "requirements*.txt",
+        "pyproject.toml",
+        "setup.py"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P04.5",
+      "name": "sbom-in-ci",
+      "pcre": "(syft|cyclonedx|spdx|sbom)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI04",
+      "description": "Trusted registry (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.sh",
+        "*.yml",
+        "*.yaml",
+        "Dockerfile",
+        "requirements*.txt",
+        "pyproject.toml",
+        "setup.py"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P04.6",
+      "name": "trusted-registry",
+      "pcre": "(index-url|extra-index-url|artifactory|jfrog|verdaccio)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI04",
+      "description": "Hash-pinned install (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.sh",
+        "*.yml",
+        "*.yaml",
+        "Dockerfile",
+        "requirements*.txt",
+        "pyproject.toml",
+        "setup.py"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P04.7",
+      "name": "hash-pinned-install",
+      "pcre": "(--require-hashes|integrity\\s*:\\s*sha)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI05",
+      "description": "Document ingestion calls",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P05.1",
+      "name": "document-ingestion-calls",
+      "pcre": "(loader\\.load\\(|ingest_document\\(|add_documents\\(|index_document\\(|UnstructuredFileLoader|PyPDFLoader|DirectoryLoader)",
+      "requires_nearby": {
+        "rules": [
+          "P05.2",
+          "P05.3"
+        ],
+        "scope": "module"
+      },
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI05",
+      "description": "Access control (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P05.2",
+      "name": "access-control",
+      "pcre": "(acl_filter|access_check|permitted_docs|filter\\s*=.*tenant|namespace\\s*=.*tenant)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI05",
+      "description": "Tenant filter on search (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P05.3",
+      "name": "tenant-filter-on-search",
+      "pcre": "(similarity_search|vector_search)[^)]{0,100}(filter|namespace|where)[^)]{0,40}tenant",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI05",
+      "description": "Integrity check on docs (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P05.4",
+      "name": "integrity-check-on-docs",
+      "pcre": "(hashlib|sha256[^\\n]{0,40}doc|integrity[._-]check|verify[._-]document)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI05",
+      "description": "Chunk size limits (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P05.5",
+      "name": "chunk-size-limits",
+      "pcre": "(max_chunk_size|chunk_size\\s*=|max_doc_size|content_limit)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI05",
+      "description": "Path traversal risk",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P05.6",
+      "name": "path-traversal-risk",
+      "pcre": "(open\\(.*\\.\\.|os\\.path\\.join\\(.*request\\.|Path\\(.*request\\.)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI06",
+      "description": "Insecure MCP transport (FAIL)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.json",
+        "*.yaml",
+        "*.toml",
+        "*.py",
+        "*.ts",
+        "mcp.json",
+        "claude_desktop_config.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P06.1",
+      "name": "insecure-mcp-transport",
+      "pcre": "(\"url\"\\s*:\\s*\"http://|transport[^\\n]{0,20}http://|\"command\"[^}]{0,200}--http(?!s))",
+      "signal": "fail"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI06",
+      "description": "MCP auth (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.json",
+        "*.yaml",
+        "*.toml",
+        "*.py",
+        "*.ts",
+        "mcp.json",
+        "claude_desktop_config.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P06.2",
+      "name": "mcp-auth",
+      "pcre": "(mcp[^\\n]{0,30}(api_key|auth|bearer)|x-api-key[^\\n]{0,20}mcp)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI06",
+      "description": "Tool schema validation (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.json",
+        "*.yaml",
+        "*.toml",
+        "*.py",
+        "*.ts",
+        "mcp.json",
+        "claude_desktop_config.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P06.3",
+      "name": "tool-schema-validation",
+      "pcre": "(jsonschema|pydantic[^\\n]{0,30}validate|zod|schema[^\\n]{0,20}tool|input_schema)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI06",
+      "description": "Wildcard tool perms (WARN)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.json",
+        "*.yaml",
+        "*.toml",
+        "*.py",
+        "*.ts",
+        "mcp.json",
+        "claude_desktop_config.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P06.4",
+      "name": "wildcard-tool-perms",
+      "pcre": "(tools\\s*[:=]\\s*\\*|\"tools\"\\s*:\\s*\"\\*\"|allow_all_tools)",
+      "signal": "warn"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI06",
+      "description": "uvicorn bind-all + no auth",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.json",
+        "*.yaml",
+        "*.toml",
+        "*.py",
+        "*.ts",
+        "mcp.json",
+        "claude_desktop_config.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P06.5",
+      "name": "uvicorn-bind-all-no-auth",
+      "pcre": "uvicorn\\.run\\([^)]*host\\s*=\\s*[\"']0\\.0\\.0\\.0",
+      "requires_nearby": {
+        "absent": true,
+        "rule": "P06.2",
+        "scope": "module"
+      },
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI07",
+      "description": "TTL config",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P07.1",
+      "name": "ttl-config",
+      "notes": "Absence of P07.1-P07.4 in a multi-tenant or PII-handling repo = WARN.",
+      "pcre": "(ttl\\s*=|expires_in\\s*=|max_age\\s*=|RETENTION_DAYS|DATA_TTL|HISTORY_TTL)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI07",
+      "description": "Session cleanup (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P07.2",
+      "name": "session-cleanup",
+      "pcre": "(delete_session|clear_history|purge_conversation|clear_memory|delete_conversation)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI07",
+      "description": "Vector delete (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P07.3",
+      "name": "vector-delete",
+      "pcre": "(delete_namespace|delete_collection|drop_index|delete_index|reset_collection)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI07",
+      "description": "Right-to-erasure (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P07.4",
+      "name": "right-to-erasure",
+      "pcre": "(gdpr_delete|erase_user_data|handle_deletion|right_to_erasure|forget_user)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI08",
+      "description": "DPIA / privacy assessment doc",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.md",
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "docs/**",
+        "appsec/**",
+        "security-review/**"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P08.1",
+      "name": "dpia-privacy-assessment-doc",
+      "notes": "Absence of P08.1-P08.6 in a production GenAI service = WARN; absence in a high-risk EU AI Act use case = FAIL.",
+      "pcre": "(DPIA|PIA|privacy_assessment|privacy[._-]impact)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI08",
+      "description": "Data processing agreement ref",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.md",
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "docs/**",
+        "appsec/**",
+        "security-review/**"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P08.2",
+      "name": "data-processing-agreement-ref",
+      "pcre": "(data_processing_agreement|DPA|sub_processor)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI08",
+      "description": "Consent capture",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.md",
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "docs/**",
+        "appsec/**",
+        "security-review/**"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P08.3",
+      "name": "consent-capture",
+      "pcre": "(consent_capture|capture_consent|consent_record|lawful_basis)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI08",
+      "description": "EU AI Act annotation",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.md",
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "docs/**",
+        "appsec/**",
+        "security-review/**"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P08.4",
+      "name": "eu-ai-act-annotation",
+      "pcre": "(ai_act_risk_level|high_risk_ai|limited_risk|minimal_risk)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI08",
+      "description": "Audit logging",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.md",
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "docs/**",
+        "appsec/**",
+        "security-review/**"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P08.5",
+      "name": "audit-logging",
+      "pcre": "(audit_log|decision_log|audit_trail)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI08",
+      "description": "Do-not-track honored",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.md",
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.json",
+        "docs/**",
+        "appsec/**",
+        "security-review/**"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P08.6",
+      "name": "do-not-track-honored",
+      "pcre": "(do_not_track|opt_out|DNT)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI09",
+      "description": "EXIF / metadata strip",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "multimodal",
+      "id": "P09.1",
+      "name": "exif-metadata-strip",
+      "pcre": "(strip_exif|remove_metadata|PIL\\.Image|exifread|piexif)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI09",
+      "description": "Image PII / moderation",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "multimodal",
+      "id": "P09.2",
+      "name": "image-pii-moderation",
+      "pcre": "(detect_pii_in_image|content_filter|nsfw_detect|moderate_image)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI09",
+      "description": "MIME / file type validation",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "multimodal",
+      "id": "P09.3",
+      "name": "mime-file-type-validation",
+      "pcre": "(allowed_types|mime_type_check|magic\\.from_buffer|filetype\\.guess)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI09",
+      "description": "Size limit on upload",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "multimodal",
+      "id": "P09.4",
+      "name": "size-limit-on-upload",
+      "pcre": "(max_upload_size|MAX_FILE_SIZE|content[._-]length[._-]limit)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI09",
+      "description": "Steganography detection",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "multimodal",
+      "id": "P09.5",
+      "name": "steganography-detection",
+      "notes": "Absence = note only (advanced control); P09.1-P09.4 absence in a multimodal pipeline = WARN.",
+      "pcre": "(stego|steganography|stegano|lsb_detect)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI10",
+      "description": "Membership inference test",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ipynb",
+        "*.r"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "synthetic_data",
+      "id": "P10.1",
+      "name": "membership-inference-test",
+      "notes": "Synthetic data pipeline without any of P10.1/P10.2/P10.4 = FAIL.",
+      "pcre": "(membership_inference|mia_test|mia_attack)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI10",
+      "description": "Anonymity validation",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ipynb",
+        "*.r"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "synthetic_data",
+      "id": "P10.2",
+      "name": "anonymity-validation",
+      "pcre": "(k_anonymity|l_diversity|t_closeness|anonymization_check)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI10",
+      "description": "Privacy-preserving lib",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ipynb",
+        "*.r"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "synthetic_data",
+      "id": "P10.3",
+      "name": "privacy-preserving-lib",
+      "pcre": "(SDV|gretel|synthetic_data_vault|smartnoise)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI10",
+      "description": "DP noise in generation",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ipynb",
+        "*.r"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "synthetic_data",
+      "id": "P10.4",
+      "name": "dp-noise-in-generation",
+      "pcre": "(epsilon\\s*=|noise_multiplier\\s*=|dp_noise|laplace_noise|gaussian_noise)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI10",
+      "description": "Re-identification risk note",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ipynb",
+        "*.r"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "synthetic_data",
+      "id": "P10.5",
+      "name": "re-identification-risk-note",
+      "pcre": "(reidentification_risk|reid_check|disclosure_risk)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI11",
+      "description": "Vector query call sites",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P11.1",
+      "name": "vector-query-call-sites",
+      "pcre": "(similarity_search|max_marginal_relevance_search|as_retriever|(vectorstore|vector_store|retriever|index|collection|client)\\.(query|search)\\()",
+      "requires_nearby": {
+        "lines": 15,
+        "rule": "P11.2"
+      },
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI11",
+      "description": "Tenant filter present (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P11.2",
+      "name": "tenant-filter-present",
+      "pcre": "(namespace\\s*=[^,)]{0,40}tenant|collection[^=]{0,20}=\\s*f?[\"'][^\"']*\\{?tenant|filter\\s*=\\s*\\{[^}]*tenant)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI11",
+      "description": "Tenant verification (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P11.3",
+      "name": "tenant-verification",
+      "pcre": "(tenant_id\\s*in\\s*session|verify_tenant|assert_tenant|check_tenant|require_tenant)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI11",
+      "description": "Cross-tenant cache risk",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P11.4",
+      "name": "cross-tenant-cache-risk",
+      "pcre": "(global[._-]cache|shared[._-]prompt[._-]cache|@cache(?!\\s*\\(.*tenant))",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI12",
+      "description": "Raw LLM-output execution (FAIL)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P12.1",
+      "name": "raw-llm-output-execution",
+      "pcre": "execute\\([^)]*(llm|response|completion|output|generated)",
+      "signal": "fail"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI12",
+      "description": "LangChain SQL agent (FAIL signal)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P12.2",
+      "name": "langchain-sql-agent",
+      "pcre": "(SQLDatabaseChain|create_sql_agent|SQLDatabaseToolkit)",
+      "signal": "fail"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI12",
+      "description": "Parameterized query (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P12.3",
+      "name": "parameterized-query",
+      "pcre": "(cursor\\.execute\\([^)]+,\\s*[\\[\\(]|prepareStatement|bindValue|:param)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI12",
+      "description": "Read-only conn (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P12.4",
+      "name": "read-only-conn",
+      "pcre": "(read_only\\s*=\\s*True|readonly\\s*=\\s*True|read_only_connection|default_transaction_read_only)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI12",
+      "description": "Query validator (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P12.5",
+      "name": "query-validator",
+      "pcre": "(validate_query|query_validator|safe_sql|sql_guard|sanitize_sql)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI12",
+      "description": "DDL from agent (FAIL)",
+      "exclude_globs": [
+        "**/migrations/**",
+        "**/fixtures/**",
+        "**/tests/**",
+        "**/test/**"
+      ],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P12.6",
+      "name": "ddl-from-agent",
+      "notes": "DDL in migrations/fixtures/tests is benign — excluded from FAIL.",
+      "pcre": "\\b(DROP\\s+TABLE|TRUNCATE|ALTER\\s+TABLE|DELETE\\s+FROM)\\b",
+      "signal": "fail"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI12",
+      "description": "Row limit (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P12.7",
+      "name": "row-limit",
+      "pcre": "(LIMIT\\s+\\d+|max_rows\\s*=|fetch_limit\\s*=)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "low",
+      "control": "DSGAI13",
+      "description": "Vector store auth configured",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P13.1",
+      "name": "vector-store-auth-configured",
+      "pcre": "(CHROMA_SERVER_AUTH|QDRANT_API_KEY|PINECONE_API_KEY|WEAVIATE_API_KEY|MILVUS_TOKEN|vectorstore[^\\n]{0,30}api_key)",
+      "signal": "info"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "medium",
+      "control": "DSGAI13",
+      "description": "TLS endpoints (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P13.2",
+      "name": "tls-endpoints",
+      "pcre": "(https://[^\"']*?(qdrant|weaviate|pinecone|chroma)|ssl\\s*=\\s*True|tls\\s*=\\s*True|grpcs://)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "low",
+      "control": "DSGAI13",
+      "description": "Insecure binding (WARN)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P13.3",
+      "name": "insecure-binding",
+      "pcre": "(host[^\\n]{0,15}0\\.0\\.0\\.0|ALLOW_RESET\\s*=\\s*True|chroma[^\\n]{0,30}http://localhost|qdrant[^\\n]{0,30}http://localhost)",
+      "signal": "warn"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "high",
+      "control": "DSGAI13",
+      "description": "Hardcoded vector token (FAIL)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go",
+        "*.yaml",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P13.4",
+      "name": "hardcoded-vector-token",
+      "pcre": "(QDRANT_API_KEY|PINECONE_API_KEY|WEAVIATE_API_KEY)\\s*=\\s*[\"'][A-Za-z0-9_\\-]{16,}",
+      "signal": "fail"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "medium",
+      "control": "DSGAI14",
+      "description": "Prompt logging disabled (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.yaml",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P14.1",
+      "name": "prompt-logging-disabled",
+      "pcre": "(log_prompts\\s*=\\s*False|capture_content\\s*=\\s*False|OTEL_LOG_PROMPTS[^\\n]{0,10}false|log_completions\\s*=\\s*False)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "low",
+      "control": "DSGAI14",
+      "description": "Prompt logging enabled (WARN)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.yaml",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P14.2",
+      "name": "prompt-logging-enabled",
+      "pcre": "(log_prompts\\s*=\\s*True|capture_content\\s*=\\s*True|log_full_response\\s*=\\s*True)",
+      "signal": "warn"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "medium",
+      "control": "DSGAI14",
+      "description": "PII redaction in log (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.yaml",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P14.3",
+      "name": "pii-redaction-in-log",
+      "pcre": "(redact_pii|mask_sensitive|sanitize_log|scrub[^\\n]{0,10}log|log[^\\n]{0,10}redact)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "low",
+      "control": "DSGAI14",
+      "description": "Raw response print (WARN)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.yaml",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P14.4",
+      "name": "raw-response-print",
+      "notes": "May contain inline PII in the format string — treated as VALUE-BEARING.",
+      "pcre": "(print\\((response|completion)|console\\.log\\((response|completion)|logger\\.(debug|info)\\([^)]*(response|completion))",
+      "signal": "warn"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "low",
+      "control": "DSGAI14",
+      "description": "LangSmith / Langfuse config",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.yaml",
+        "*.env*"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P14.5",
+      "name": "langsmith-langfuse-config",
+      "pcre": "(LANGSMITH_API_KEY|LANGFUSE_PUBLIC_KEY|langfuse[^\\n]{0,20}init|langsmith[^\\n]{0,20}trace)",
+      "signal": "info"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "high",
+      "control": "DSGAI15",
+      "description": "Secret in system prompt (FAIL)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P15.1",
+      "name": "secret-in-system-prompt",
+      "pcre": "(system_prompt|system_message)\\s*=\\s*[\"'][^\"']*(api_key|password|secret|token|sk-)",
+      "signal": "fail"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "medium",
+      "control": "DSGAI15",
+      "description": "Context size limit (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P15.2",
+      "name": "context-size-limit",
+      "pcre": "(max_context_length\\s*=|context_window_limit\\s*=|max_context_tokens\\s*=|truncate_context)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "medium",
+      "control": "DSGAI15",
+      "description": "Prompt from config (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P15.3",
+      "name": "prompt-from-config",
+      "pcre": "(system_prompt|system_message)[^\\n]{0,40}(config|vault|os\\.environ|getenv|secret_manager)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "value_bearing",
+      "confidence": "low",
+      "control": "DSGAI15",
+      "description": "Over-fetch aggregation (WARN)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P15.4",
+      "name": "over-fetch-aggregation",
+      "pcre": "(customer_360|user_profile_all|fetch_all_user_data|get_full_profile|select\\s*\\*\\s*from\\s+users)",
+      "signal": "warn"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI16",
+      "description": "AI-ignore file present (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        ".copilotignore",
+        ".aiignore",
+        ".cursorignore",
+        ".continueignore",
+        ".codeiumignore",
+        ".vscode/settings.json",
+        "cursor.json",
+        "continue.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P16.1",
+      "name": "ai-ignore-file-present",
+      "notes": "Filename-existence check (not content grep). Absence in a repo with .env or secrets/ = WARN.",
+      "pcre": "(\\.(copilot|ai|cursor|continue|codeium)ignore)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI16",
+      "description": "Sensitive paths excluded",
+      "exclude_globs": [],
+      "file_globs": [
+        ".copilotignore",
+        ".aiignore",
+        ".cursorignore",
+        ".continueignore",
+        ".codeiumignore",
+        ".vscode/settings.json",
+        "cursor.json",
+        "continue.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P16.2",
+      "name": "sensitive-paths-excluded",
+      "notes": "Matched inside any AI-ignore file.",
+      "pcre": "(\\.env|secrets|credentials|\\.aws|\\.ssh|private_key)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI16",
+      "description": "Telemetry off (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        ".copilotignore",
+        ".aiignore",
+        ".cursorignore",
+        ".continueignore",
+        ".codeiumignore",
+        ".vscode/settings.json",
+        "cursor.json",
+        "continue.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P16.3",
+      "name": "telemetry-off",
+      "pcre": "(telemetry[^\\n]{0,10}(off|false|disabled)|share_data\\s*[:=]\\s*false)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI16",
+      "description": "Context scope limits",
+      "exclude_globs": [],
+      "file_globs": [
+        ".copilotignore",
+        ".aiignore",
+        ".cursorignore",
+        ".continueignore",
+        ".codeiumignore",
+        ".vscode/settings.json",
+        "cursor.json",
+        "continue.json"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P16.4",
+      "name": "context-scope-limits",
+      "pcre": "(contextWindow|ignorePaths|excludeFiles|maxContextLines)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI17",
+      "description": "Circuit breaker (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P17.1",
+      "name": "circuit-breaker",
+      "notes": "LLM-calling module with none of P17.1-P17.5 = WARN.",
+      "pcre": "(CircuitBreaker|@circuit|tenacity|resilience4j|pybreaker)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI17",
+      "description": "Retry with backoff (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P17.2",
+      "name": "retry-with-backoff",
+      "pcre": "(retry\\s*\\(|@retry|exponential_backoff|max_retries\\s*=|backoff_factor\\s*=|retry_with_backoff)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI17",
+      "description": "Timeout on LLM call (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P17.3",
+      "name": "timeout-on-llm-call",
+      "pcre": "(timeout\\s*=\\s*\\d|request_timeout\\s*=|httpx\\.[^\\n]{0,30}timeout)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI17",
+      "description": "Rate limit incoming (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P17.4",
+      "name": "rate-limit-incoming",
+      "pcre": "(rate_limit|@throttle|RateLimiter|slowapi|flask_limiter|express-rate-limit)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI17",
+      "description": "Fallback response (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P17.5",
+      "name": "fallback-response",
+      "pcre": "(fallback_response|default_response|graceful_degradation|on_failure_return)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI17",
+      "description": "Unbounded retry (FAIL)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P17.6",
+      "name": "unbounded-retry",
+      "pcre": "while\\s+True[^}]{0,200}(generate|complete|chat)|retry\\s*\\(\\s*\\)",
+      "signal": "fail"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI18",
+      "description": "Output guardrails (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P18.1",
+      "name": "output-guardrails",
+      "pcre": "(guardrails|nemo[._-]guardrails|llm_guard|output_guard|filter_output|sanitize_response|moderation)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI18",
+      "description": "PII detection on output",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P18.2",
+      "name": "pii-detection-on-output",
+      "pcre": "(detect_pii[^\\n]{0,20}output|output[^\\n]{0,20}detect_pii|presidio[^\\n]{0,20}output|scan_output)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI18",
+      "description": "Logprobs exposed (WARN)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P18.3",
+      "name": "logprobs-exposed",
+      "pcre": "(logprobs\\s*=\\s*True|return_logprobs\\s*=\\s*True|include_logprobs\\s*=\\s*True|top_logprobs\\s*=\\s*\\d)",
+      "signal": "warn"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI18",
+      "description": "LLM call sites (count)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P18.4",
+      "name": "llm-call-sites",
+      "pcre": "(\\.chat\\.completions\\.create|client\\.messages\\.create|openai\\.ChatCompletion|anthropic\\.messages|generate_content|chat\\.complete)",
+      "requires_nearby": {
+        "lines": 10,
+        "rule": "P18.5"
+      },
+      "signal": "count"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI18",
+      "description": "max_tokens set (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P18.5",
+      "name": "max-tokens-set",
+      "pcre": "max_tokens\\s*=\\s*\\d+",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI19",
+      "description": "Anonymization before export",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ipynb",
+        "*.ts",
+        "*.yaml"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "labeling",
+      "id": "P19.1",
+      "name": "anonymization-before-export",
+      "pcre": "(anonymize_for_labeling|pseudonymize|redact_before_export|mask_for_labeling)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI19",
+      "description": "Labeling SDK auth from vault",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ipynb",
+        "*.ts",
+        "*.yaml"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "labeling",
+      "id": "P19.2",
+      "name": "labeling-sdk-auth-from-vault",
+      "pcre": "(label_studio[^\\n]{0,30}vault|labelbox[^\\n]{0,30}secret|scale_api_key[^\\n]{0,30}(vault|env))",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI19",
+      "description": "Labeling access audit",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ipynb",
+        "*.ts",
+        "*.yaml"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "labeling",
+      "id": "P19.3",
+      "name": "labeling-access-audit",
+      "pcre": "(label[._-]audit|labeling_access_log|annotator_audit)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI19",
+      "description": "Re-identification post-label",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ipynb",
+        "*.ts",
+        "*.yaml"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "gated_on": "labeling",
+      "id": "P19.4",
+      "name": "re-identification-post-label",
+      "pcre": "(reid_check_post_label|labeling_reidentification|post_label_validation)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI20",
+      "description": "API auth required (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P20.1",
+      "name": "api-auth-required",
+      "pcre": "(api_key[^\\n]{0,15}(verify|validate)|bearer_token[^\\n]{0,15}validate|Authorization[^\\n]{0,15}required|@require_auth|auth_middleware|Depends\\(.*auth)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI20",
+      "description": "Rate limit (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P20.2",
+      "name": "rate-limit",
+      "pcre": "(rate_limit|RateLimiter|@throttle|slowapi|flask_limiter|express-rate-limit|@limiter\\.limit)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI20",
+      "description": "Input length validation (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P20.3",
+      "name": "input-length-validation",
+      "pcre": "(max_input_length\\s*=|max_input_tokens\\s*=|input[._-]truncat|validate[._-]length)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI20",
+      "description": "Prompt injection detect (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P20.4",
+      "name": "prompt-injection-detect",
+      "pcre": "(prompt_injection_detect|detect_injection|llm_guard|rebuff|input_guard|nemo[._-]guard)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI20",
+      "description": "Inference endpoint (count)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt",
+        "*.go"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P20.5",
+      "name": "inference-endpoint",
+      "pcre": "(@app\\.(post|get)\\([\"'][^\"']*(chat|generate|completion|infer|predict)|@router\\.(post|get)\\([\"'][^\"']*(chat|generate|completion|infer|predict)|app\\.(post|get)\\([\"'][^\"']*(chat|generate|completion|infer|predict))",
+      "requires_nearby": {
+        "lines": 15,
+        "rules": [
+          "P20.1",
+          "P20.2"
+        ]
+      },
+      "signal": "count"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI21",
+      "description": "Read-only retrieval (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P21.1",
+      "name": "read-only-retrieval",
+      "pcre": "(read_only\\s*=\\s*True|readonly\\s*=\\s*True|HttpMethod\\.GET|http_method[^\\n]{0,10}get)",
+      "signal": "pass_signal"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI21",
+      "description": "KB write from agent (WARN)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P21.2",
+      "name": "kb-write-from-agent",
+      "notes": "P21.2 in an agent module without P21.3 = WARN.",
+      "pcre": "(chromadb|qdrant|pinecone|weaviate|knowledge_base|kb_client)[^\\n]{0,40}\\.(upsert|insert|update|delete|add_documents|write)\\(",
+      "signal": "warn"
+    },
+    {
+      "classification": "structural",
+      "confidence": "low",
+      "control": "DSGAI21",
+      "description": "Write content validation",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P21.3",
+      "name": "write-content-validation",
+      "pcre": "(validate_content|sanitize[._-]write|verify[._-]knowledge|content[._-]check[._-]write)",
+      "signal": "info"
+    },
+    {
+      "classification": "structural",
+      "confidence": "medium",
+      "control": "DSGAI21",
+      "description": "KB versioning (PASS)",
+      "exclude_globs": [],
+      "file_globs": [
+        "*.py",
+        "*.ts",
+        "*.java",
+        "*.kt"
+      ],
+      "framework": "dsgai-2026-v1.0",
+      "id": "P21.4",
+      "name": "kb-versioning",
+      "pcre": "(versioned\\s*=\\s*True|audit_writes|version_id|kb_audit|knowledge[._-]audit)",
+      "signal": "pass_signal"
+    }
+  ],
+  "ruleset_version": "0.3.0"
+}
diff --git a/dsgai_scanner_tool/rules/dsgai-rules.yaml b/dsgai_scanner_tool/rules/dsgai-rules.yaml
new file mode 100644
index 0000000..6816ea4
--- /dev/null
+++ b/dsgai_scanner_tool/rules/dsgai-rules.yaml
@@ -0,0 +1,1202 @@
+# DSGAI scanner detection rules — source of truth.
+# Generated once from dsgai_scanner_tool.md Step 2 by build/generate_rules.py,
+# then hand-maintained. Validated by rules/rules.schema.json (see rules/README.md).
+ruleset_version: '0.3.0'
+framework: 'dsgai-2026-v1.0'
+rules:
+  - id: P01.1
+    control: DSGAI01
+    name: pii-scrubbing-imports
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(anonymize|redact|scrub_pii|piiDetect|mask_pii|presidio)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'PII scrubbing imports'
+  - id: P01.2
+    control: DSGAI01
+    name: no-train-flags
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(allow_training|training_opt_out|X-Training-Data|no_train)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'No-train flags'
+  - id: P01.3
+    control: DSGAI01
+    name: output-pii-filter
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(filter_pii|remove_pii|output_sanitize|output_filter)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Output PII filter'
+  - id: P01.4
+    control: DSGAI01
+    name: differential-privacy-lib
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(opacus|tensorflow\.privacy|dp-accounting|differential\.privacy)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Differential privacy lib'
+  - id: P01.5
+    control: DSGAI01
+    name: consent-gdpr-check
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(consent_capture|gdpr|ccpa|data_subject|lawful_basis)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Consent / GDPR check'
+  - id: P02.1
+    control: DSGAI02
+    name: hardcoded-llm-api-key
+    classification: value_bearing
+    signal: fail
+    confidence: high
+    pcre: '(?i)(OPENAI_API_KEY|openai[._-]?api[._-]?key)\s*[:=]\s*["'']sk-[A-Za-z0-9_\-]{20,}'
+    file_globs: ['*.py', '*.ts', '*.js', '*.java', '*.kt', '*.go', '*.env*', '*.yaml', '*.yml', '*.json', '*.toml', '*.cfg']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Hardcoded LLM API key (OpenAI)'
+  - id: P02.2
+    control: DSGAI02
+    name: hardcoded-llm-api-key
+    classification: value_bearing
+    signal: fail
+    confidence: high
+    pcre: '(?i)(ANTHROPIC_API_KEY|anthropic[._-]?api[._-]?key)\s*[:=]\s*["'']sk-ant-[A-Za-z0-9_\-]{20,}'
+    file_globs: ['*.py', '*.ts', '*.js', '*.java', '*.kt', '*.go', '*.env*', '*.yaml', '*.yml', '*.json', '*.toml', '*.cfg']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Hardcoded LLM API key (Anthropic)'
+  - id: P02.3
+    control: DSGAI02
+    name: hardcoded-cohere-google-hf-tokens
+    classification: value_bearing
+    signal: fail
+    confidence: high
+    pcre: '(?i)(COHERE_API_KEY|GOOGLE_API_KEY|HF_TOKEN|HUGGINGFACE_TOKEN)\s*[:=]\s*["''][A-Za-z0-9_\-]{20,}'
+    file_globs: ['*.py', '*.ts', '*.js', '*.java', '*.kt', '*.go', '*.env*', '*.yaml', '*.yml', '*.json', '*.toml', '*.cfg']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Hardcoded Cohere/Google/HF tokens'
+  - id: P02.4
+    control: DSGAI02
+    name: hardcoded-aws-creds
+    classification: value_bearing
+    signal: fail
+    confidence: high
+    pcre: '(AWS_ACCESS_KEY_ID|aws_access_key_id)\s*[:=]\s*["''][A-Z0-9]{16,}'
+    file_globs: ['*.py', '*.ts', '*.js', '*.java', '*.kt', '*.go', '*.env*', '*.yaml', '*.yml', '*.json', '*.toml', '*.cfg']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Hardcoded AWS creds'
+  - id: P02.5
+    control: DSGAI02
+    name: hardcoded-azure-gcp-cred
+    classification: value_bearing
+    signal: fail
+    confidence: high
+    pcre: '(AZURE_OPENAI_KEY|GCP_SERVICE_ACCOUNT_KEY)\s*[:=]\s*["''][A-Za-z0-9_\-]{16,}'
+    file_globs: ['*.py', '*.ts', '*.js', '*.java', '*.kt', '*.go', '*.env*', '*.yaml', '*.yml', '*.json', '*.toml', '*.cfg']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Hardcoded Azure/GCP cred'
+  - id: P02.6
+    control: DSGAI02
+    name: wildcard-token-scope
+    classification: value_bearing
+    signal: warn
+    confidence: low
+    pcre: '("scope"\s*:\s*"\*|permissions[^\n]{0,30}\*|scope[^\n]{0,20}admin)'
+    file_globs: ['*.py', '*.ts', '*.js', '*.java', '*.kt', '*.go', '*.env*', '*.yaml', '*.yml', '*.json', '*.toml', '*.cfg']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Wildcard token scope (warn)'
+  - id: P02.7
+    control: DSGAI02
+    name: vault-secrets-manager
+    classification: value_bearing
+    signal: pass_signal
+    confidence: medium
+    pcre: '(hvac\.Client|hashicorp/vault|VAULT_(ADDR|TOKEN|NAMESPACE)|secretsmanager\.|GetSecretValue|SecretManagerServiceClient|azure[._-]keyvault|@aws-sdk/client-secrets-manager)'
+    file_globs: ['*.py', '*.ts', '*.js', '*.java', '*.kt', '*.go', '*.env*', '*.yaml', '*.yml', '*.json', '*.toml', '*.cfg']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Vault/secrets-manager (PASS)'
+  - id: P02.8
+    control: DSGAI02
+    name: tool-call-signing
+    classification: value_bearing
+    signal: pass_signal
+    confidence: medium
+    pcre: '(hmac|sign_request|verify_signature|tool_auth|mtls|mutual_tls)'
+    file_globs: ['*.py', '*.ts', '*.js', '*.java', '*.kt', '*.go', '*.env*', '*.yaml', '*.yml', '*.json', '*.toml', '*.cfg']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Tool-call signing (PASS)'
+  - id: P03.1
+    control: DSGAI03
+    name: third-party-llm-endpoints
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(api\.openai\.com|api\.anthropic\.com|generativelanguage\.googleapis\.com|api\.cohere\.com|api\.together\.xyz|api\.mistral\.ai|api\.groq\.com)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.go', '*.yaml', '*.json', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Third-party LLM endpoints'
+  - id: P03.2
+    control: DSGAI03
+    name: internal-llm-gateway
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(llm[._-]?gateway|ai[._-]?proxy|model[._-]?gateway|llm[._-]?proxy)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.go', '*.yaml', '*.json', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Internal LLM gateway (PASS)'
+  - id: P03.3
+    control: DSGAI03
+    name: dlp-classification-check
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(dlp|data_classification|classify_data|sensitivity_check)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.go', '*.yaml', '*.json', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'DLP / classification check'
+  - id: P04.1
+    control: DSGAI04
+    name: unsafe-pickle
+    classification: structural
+    signal: fail
+    confidence: medium
+    pcre: 'torch\.load\s*\('
+    file_globs: ['*.py', '*.sh', '*.yml', '*.yaml', 'Dockerfile', 'requirements*.txt', 'pyproject.toml', 'setup.py']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Unsafe pickle (FAIL)'
+    subtract: ['P04.2']
+  - id: P04.2
+    control: DSGAI04
+    name: safe-pickle
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: 'torch\.load\s*\([^)]*weights_only\s*=\s*True'
+    file_globs: ['*.py', '*.sh', '*.yml', '*.yaml', 'Dockerfile', 'requirements*.txt', 'pyproject.toml', 'setup.py']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Safe pickle (PASS counter)'
+  - id: P04.3
+    control: DSGAI04
+    name: artifact-verification
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(sha256|verify_signature|model_hash|check_integrity)'
+    file_globs: ['*.py', '*.sh', '*.yml', '*.yaml', 'Dockerfile', 'requirements*.txt', 'pyproject.toml', 'setup.py']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Artifact verification (PASS)'
+  - id: P04.4
+    control: DSGAI04
+    name: unpinned-ml-deps
+    classification: structural
+    signal: warn
+    confidence: low
+    pcre: '^(torch|transformers|tensorflow|langchain|openai|anthropic|llama-index)\s*(>=|~=|\^|>|<|<=|\*|latest)'
+    file_globs: ['*.py', '*.sh', '*.yml', '*.yaml', 'Dockerfile', 'requirements*.txt', 'pyproject.toml', 'setup.py']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Unpinned ML deps (WARN)'
+  - id: P04.5
+    control: DSGAI04
+    name: sbom-in-ci
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(syft|cyclonedx|spdx|sbom)'
+    file_globs: ['*.py', '*.sh', '*.yml', '*.yaml', 'Dockerfile', 'requirements*.txt', 'pyproject.toml', 'setup.py']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'SBOM in CI (PASS)'
+  - id: P04.6
+    control: DSGAI04
+    name: trusted-registry
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(index-url|extra-index-url|artifactory|jfrog|verdaccio)'
+    file_globs: ['*.py', '*.sh', '*.yml', '*.yaml', 'Dockerfile', 'requirements*.txt', 'pyproject.toml', 'setup.py']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Trusted registry (PASS)'
+  - id: P04.7
+    control: DSGAI04
+    name: hash-pinned-install
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(--require-hashes|integrity\s*:\s*sha)'
+    file_globs: ['*.py', '*.sh', '*.yml', '*.yaml', 'Dockerfile', 'requirements*.txt', 'pyproject.toml', 'setup.py']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Hash-pinned install (PASS)'
+  - id: P05.1
+    control: DSGAI05
+    name: document-ingestion-calls
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(loader\.load\(|ingest_document\(|add_documents\(|index_document\(|UnstructuredFileLoader|PyPDFLoader|DirectoryLoader)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Document ingestion calls'
+    requires_nearby: {rules: ['P05.2', 'P05.3'], scope: module}
+  - id: P05.2
+    control: DSGAI05
+    name: access-control
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(acl_filter|access_check|permitted_docs|filter\s*=.*tenant|namespace\s*=.*tenant)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Access control (PASS)'
+  - id: P05.3
+    control: DSGAI05
+    name: tenant-filter-on-search
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(similarity_search|vector_search)[^)]{0,100}(filter|namespace|where)[^)]{0,40}tenant'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Tenant filter on search (PASS)'
+  - id: P05.4
+    control: DSGAI05
+    name: integrity-check-on-docs
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(hashlib|sha256[^\n]{0,40}doc|integrity[._-]check|verify[._-]document)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Integrity check on docs (PASS)'
+  - id: P05.5
+    control: DSGAI05
+    name: chunk-size-limits
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(max_chunk_size|chunk_size\s*=|max_doc_size|content_limit)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Chunk size limits (PASS)'
+  - id: P05.6
+    control: DSGAI05
+    name: path-traversal-risk
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(open\(.*\.\.|os\.path\.join\(.*request\.|Path\(.*request\.)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Path traversal risk'
+  - id: P06.1
+    control: DSGAI06
+    name: insecure-mcp-transport
+    classification: structural
+    signal: fail
+    confidence: medium
+    pcre: '("url"\s*:\s*"http://|transport[^\n]{0,20}http://|"command"[^}]{0,200}--http(?!s))'
+    file_globs: ['*.json', '*.yaml', '*.toml', '*.py', '*.ts', 'mcp.json', 'claude_desktop_config.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Insecure MCP transport (FAIL)'
+  - id: P06.2
+    control: DSGAI06
+    name: mcp-auth
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(mcp[^\n]{0,30}(api_key|auth|bearer)|x-api-key[^\n]{0,20}mcp)'
+    file_globs: ['*.json', '*.yaml', '*.toml', '*.py', '*.ts', 'mcp.json', 'claude_desktop_config.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'MCP auth (PASS)'
+  - id: P06.3
+    control: DSGAI06
+    name: tool-schema-validation
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(jsonschema|pydantic[^\n]{0,30}validate|zod|schema[^\n]{0,20}tool|input_schema)'
+    file_globs: ['*.json', '*.yaml', '*.toml', '*.py', '*.ts', 'mcp.json', 'claude_desktop_config.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Tool schema validation (PASS)'
+  - id: P06.4
+    control: DSGAI06
+    name: wildcard-tool-perms
+    classification: structural
+    signal: warn
+    confidence: low
+    pcre: '(tools\s*[:=]\s*\*|"tools"\s*:\s*"\*"|allow_all_tools)'
+    file_globs: ['*.json', '*.yaml', '*.toml', '*.py', '*.ts', 'mcp.json', 'claude_desktop_config.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Wildcard tool perms (WARN)'
+  - id: P06.5
+    control: DSGAI06
+    name: uvicorn-bind-all-no-auth
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: 'uvicorn\.run\([^)]*host\s*=\s*["'']0\.0\.0\.0'
+    file_globs: ['*.json', '*.yaml', '*.toml', '*.py', '*.ts', 'mcp.json', 'claude_desktop_config.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'uvicorn bind-all + no auth'
+    requires_nearby: {rule: P06.2, scope: module, absent: true}
+  - id: P07.1
+    control: DSGAI07
+    name: ttl-config
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(ttl\s*=|expires_in\s*=|max_age\s*=|RETENTION_DAYS|DATA_TTL|HISTORY_TTL)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'TTL config'
+    notes: 'Absence of P07.1-P07.4 in a multi-tenant or PII-handling repo = WARN.'
+  - id: P07.2
+    control: DSGAI07
+    name: session-cleanup
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(delete_session|clear_history|purge_conversation|clear_memory|delete_conversation)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Session cleanup (PASS)'
+  - id: P07.3
+    control: DSGAI07
+    name: vector-delete
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(delete_namespace|delete_collection|drop_index|delete_index|reset_collection)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Vector delete (PASS)'
+  - id: P07.4
+    control: DSGAI07
+    name: right-to-erasure
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(gdpr_delete|erase_user_data|handle_deletion|right_to_erasure|forget_user)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Right-to-erasure (PASS)'
+  - id: P08.1
+    control: DSGAI08
+    name: dpia-privacy-assessment-doc
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(DPIA|PIA|privacy_assessment|privacy[._-]impact)'
+    file_globs: ['*.md', '*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json', 'docs/**', 'appsec/**', 'security-review/**']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'DPIA / privacy assessment doc'
+    notes: 'Absence of P08.1-P08.6 in a production GenAI service = WARN; absence in a high-risk EU AI Act use case = FAIL.'
+  - id: P08.2
+    control: DSGAI08
+    name: data-processing-agreement-ref
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(data_processing_agreement|DPA|sub_processor)'
+    file_globs: ['*.md', '*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json', 'docs/**', 'appsec/**', 'security-review/**']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Data processing agreement ref'
+  - id: P08.3
+    control: DSGAI08
+    name: consent-capture
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(consent_capture|capture_consent|consent_record|lawful_basis)'
+    file_globs: ['*.md', '*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json', 'docs/**', 'appsec/**', 'security-review/**']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Consent capture'
+  - id: P08.4
+    control: DSGAI08
+    name: eu-ai-act-annotation
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(ai_act_risk_level|high_risk_ai|limited_risk|minimal_risk)'
+    file_globs: ['*.md', '*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json', 'docs/**', 'appsec/**', 'security-review/**']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'EU AI Act annotation'
+  - id: P08.5
+    control: DSGAI08
+    name: audit-logging
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(audit_log|decision_log|audit_trail)'
+    file_globs: ['*.md', '*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json', 'docs/**', 'appsec/**', 'security-review/**']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Audit logging'
+  - id: P08.6
+    control: DSGAI08
+    name: do-not-track-honored
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(do_not_track|opt_out|DNT)'
+    file_globs: ['*.md', '*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.json', 'docs/**', 'appsec/**', 'security-review/**']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Do-not-track honored'
+  - id: P09.1
+    control: DSGAI09
+    name: exif-metadata-strip
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(strip_exif|remove_metadata|PIL\.Image|exifread|piexif)'
+    file_globs: ['*.py', '*.ts', '*.java']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'EXIF / metadata strip'
+    gated_on: multimodal
+  - id: P09.2
+    control: DSGAI09
+    name: image-pii-moderation
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(detect_pii_in_image|content_filter|nsfw_detect|moderate_image)'
+    file_globs: ['*.py', '*.ts', '*.java']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Image PII / moderation'
+    gated_on: multimodal
+  - id: P09.3
+    control: DSGAI09
+    name: mime-file-type-validation
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(allowed_types|mime_type_check|magic\.from_buffer|filetype\.guess)'
+    file_globs: ['*.py', '*.ts', '*.java']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'MIME / file type validation'
+    gated_on: multimodal
+  - id: P09.4
+    control: DSGAI09
+    name: size-limit-on-upload
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(max_upload_size|MAX_FILE_SIZE|content[._-]length[._-]limit)'
+    file_globs: ['*.py', '*.ts', '*.java']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Size limit on upload'
+    gated_on: multimodal
+  - id: P09.5
+    control: DSGAI09
+    name: steganography-detection
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(stego|steganography|stegano|lsb_detect)'
+    file_globs: ['*.py', '*.ts', '*.java']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Steganography detection'
+    gated_on: multimodal
+    notes: 'Absence = note only (advanced control); P09.1-P09.4 absence in a multimodal pipeline = WARN.'
+  - id: P10.1
+    control: DSGAI10
+    name: membership-inference-test
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(membership_inference|mia_test|mia_attack)'
+    file_globs: ['*.py', '*.ipynb', '*.r']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Membership inference test'
+    gated_on: synthetic_data
+    notes: 'Synthetic data pipeline without any of P10.1/P10.2/P10.4 = FAIL.'
+  - id: P10.2
+    control: DSGAI10
+    name: anonymity-validation
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(k_anonymity|l_diversity|t_closeness|anonymization_check)'
+    file_globs: ['*.py', '*.ipynb', '*.r']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Anonymity validation'
+    gated_on: synthetic_data
+  - id: P10.3
+    control: DSGAI10
+    name: privacy-preserving-lib
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(SDV|gretel|synthetic_data_vault|smartnoise)'
+    file_globs: ['*.py', '*.ipynb', '*.r']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Privacy-preserving lib'
+    gated_on: synthetic_data
+  - id: P10.4
+    control: DSGAI10
+    name: dp-noise-in-generation
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(epsilon\s*=|noise_multiplier\s*=|dp_noise|laplace_noise|gaussian_noise)'
+    file_globs: ['*.py', '*.ipynb', '*.r']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'DP noise in generation'
+    gated_on: synthetic_data
+  - id: P10.5
+    control: DSGAI10
+    name: re-identification-risk-note
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(reidentification_risk|reid_check|disclosure_risk)'
+    file_globs: ['*.py', '*.ipynb', '*.r']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Re-identification risk note'
+    gated_on: synthetic_data
+  - id: P11.1
+    control: DSGAI11
+    name: vector-query-call-sites
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(similarity_search|max_marginal_relevance_search|as_retriever|(vectorstore|vector_store|retriever|index|collection|client)\.(query|search)\()'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Vector query call sites'
+    requires_nearby: {rule: P11.2, lines: 15}
+  - id: P11.2
+    control: DSGAI11
+    name: tenant-filter-present
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(namespace\s*=[^,)]{0,40}tenant|collection[^=]{0,20}=\s*f?["''][^"'']*\{?tenant|filter\s*=\s*\{[^}]*tenant)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Tenant filter present (PASS)'
+  - id: P11.3
+    control: DSGAI11
+    name: tenant-verification
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(tenant_id\s*in\s*session|verify_tenant|assert_tenant|check_tenant|require_tenant)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Tenant verification (PASS)'
+  - id: P11.4
+    control: DSGAI11
+    name: cross-tenant-cache-risk
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(global[._-]cache|shared[._-]prompt[._-]cache|@cache(?!\s*\(.*tenant))'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Cross-tenant cache risk'
+  - id: P12.1
+    control: DSGAI12
+    name: raw-llm-output-execution
+    classification: structural
+    signal: fail
+    confidence: medium
+    pcre: 'execute\([^)]*(llm|response|completion|output|generated)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Raw LLM-output execution (FAIL)'
+  - id: P12.2
+    control: DSGAI12
+    name: langchain-sql-agent
+    classification: structural
+    signal: fail
+    confidence: medium
+    pcre: '(SQLDatabaseChain|create_sql_agent|SQLDatabaseToolkit)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'LangChain SQL agent (FAIL signal)'
+  - id: P12.3
+    control: DSGAI12
+    name: parameterized-query
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(cursor\.execute\([^)]+,\s*[\[\(]|prepareStatement|bindValue|:param)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Parameterized query (PASS)'
+  - id: P12.4
+    control: DSGAI12
+    name: read-only-conn
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(read_only\s*=\s*True|readonly\s*=\s*True|read_only_connection|default_transaction_read_only)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Read-only conn (PASS)'
+  - id: P12.5
+    control: DSGAI12
+    name: query-validator
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(validate_query|query_validator|safe_sql|sql_guard|sanitize_sql)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Query validator (PASS)'
+  - id: P12.6
+    control: DSGAI12
+    name: ddl-from-agent
+    classification: structural
+    signal: fail
+    confidence: medium
+    pcre: '\b(DROP\s+TABLE|TRUNCATE|ALTER\s+TABLE|DELETE\s+FROM)\b'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: ['**/migrations/**', '**/fixtures/**', '**/tests/**', '**/test/**']
+    framework: 'dsgai-2026-v1.0'
+    description: 'DDL from agent (FAIL)'
+    notes: 'DDL in migrations/fixtures/tests is benign — excluded from FAIL.'
+  - id: P12.7
+    control: DSGAI12
+    name: row-limit
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(LIMIT\s+\d+|max_rows\s*=|fetch_limit\s*=)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Row limit (PASS)'
+  - id: P13.1
+    control: DSGAI13
+    name: vector-store-auth-configured
+    classification: value_bearing
+    signal: info
+    confidence: low
+    pcre: '(CHROMA_SERVER_AUTH|QDRANT_API_KEY|PINECONE_API_KEY|WEAVIATE_API_KEY|MILVUS_TOKEN|vectorstore[^\n]{0,30}api_key)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Vector store auth configured'
+  - id: P13.2
+    control: DSGAI13
+    name: tls-endpoints
+    classification: value_bearing
+    signal: pass_signal
+    confidence: medium
+    pcre: '(https://[^"'']*?(qdrant|weaviate|pinecone|chroma)|ssl\s*=\s*True|tls\s*=\s*True|grpcs://)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'TLS endpoints (PASS)'
+  - id: P13.3
+    control: DSGAI13
+    name: insecure-binding
+    classification: value_bearing
+    signal: warn
+    confidence: low
+    pcre: '(host[^\n]{0,15}0\.0\.0\.0|ALLOW_RESET\s*=\s*True|chroma[^\n]{0,30}http://localhost|qdrant[^\n]{0,30}http://localhost)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Insecure binding (WARN)'
+  - id: P13.4
+    control: DSGAI13
+    name: hardcoded-vector-token
+    classification: value_bearing
+    signal: fail
+    confidence: high
+    pcre: '(QDRANT_API_KEY|PINECONE_API_KEY|WEAVIATE_API_KEY)\s*=\s*["''][A-Za-z0-9_\-]{16,}'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go', '*.yaml', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Hardcoded vector token (FAIL)'
+  - id: P14.1
+    control: DSGAI14
+    name: prompt-logging-disabled
+    classification: value_bearing
+    signal: pass_signal
+    confidence: medium
+    pcre: '(log_prompts\s*=\s*False|capture_content\s*=\s*False|OTEL_LOG_PROMPTS[^\n]{0,10}false|log_completions\s*=\s*False)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.yaml', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Prompt logging disabled (PASS)'
+  - id: P14.2
+    control: DSGAI14
+    name: prompt-logging-enabled
+    classification: value_bearing
+    signal: warn
+    confidence: low
+    pcre: '(log_prompts\s*=\s*True|capture_content\s*=\s*True|log_full_response\s*=\s*True)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.yaml', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Prompt logging enabled (WARN)'
+  - id: P14.3
+    control: DSGAI14
+    name: pii-redaction-in-log
+    classification: value_bearing
+    signal: pass_signal
+    confidence: medium
+    pcre: '(redact_pii|mask_sensitive|sanitize_log|scrub[^\n]{0,10}log|log[^\n]{0,10}redact)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.yaml', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'PII redaction in log (PASS)'
+  - id: P14.4
+    control: DSGAI14
+    name: raw-response-print
+    classification: value_bearing
+    signal: warn
+    confidence: low
+    pcre: '(print\((response|completion)|console\.log\((response|completion)|logger\.(debug|info)\([^)]*(response|completion))'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.yaml', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Raw response print (WARN)'
+    notes: 'May contain inline PII in the format string — treated as VALUE-BEARING.'
+  - id: P14.5
+    control: DSGAI14
+    name: langsmith-langfuse-config
+    classification: value_bearing
+    signal: info
+    confidence: low
+    pcre: '(LANGSMITH_API_KEY|LANGFUSE_PUBLIC_KEY|langfuse[^\n]{0,20}init|langsmith[^\n]{0,20}trace)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.yaml', '*.env*']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'LangSmith / Langfuse config'
+  - id: P15.1
+    control: DSGAI15
+    name: secret-in-system-prompt
+    classification: value_bearing
+    signal: fail
+    confidence: high
+    pcre: '(system_prompt|system_message)\s*=\s*["''][^"'']*(api_key|password|secret|token|sk-)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Secret in system prompt (FAIL)'
+  - id: P15.2
+    control: DSGAI15
+    name: context-size-limit
+    classification: value_bearing
+    signal: pass_signal
+    confidence: medium
+    pcre: '(max_context_length\s*=|context_window_limit\s*=|max_context_tokens\s*=|truncate_context)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Context size limit (PASS)'
+  - id: P15.3
+    control: DSGAI15
+    name: prompt-from-config
+    classification: value_bearing
+    signal: pass_signal
+    confidence: medium
+    pcre: '(system_prompt|system_message)[^\n]{0,40}(config|vault|os\.environ|getenv|secret_manager)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Prompt from config (PASS)'
+  - id: P15.4
+    control: DSGAI15
+    name: over-fetch-aggregation
+    classification: value_bearing
+    signal: warn
+    confidence: low
+    pcre: '(customer_360|user_profile_all|fetch_all_user_data|get_full_profile|select\s*\*\s*from\s+users)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Over-fetch aggregation (WARN)'
+  - id: P16.1
+    control: DSGAI16
+    name: ai-ignore-file-present
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(\.(copilot|ai|cursor|continue|codeium)ignore)'
+    file_globs: ['.copilotignore', '.aiignore', '.cursorignore', '.continueignore', '.codeiumignore', '.vscode/settings.json', 'cursor.json', 'continue.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'AI-ignore file present (PASS)'
+    notes: 'Filename-existence check (not content grep). Absence in a repo with .env or secrets/ = WARN.'
+  - id: P16.2
+    control: DSGAI16
+    name: sensitive-paths-excluded
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(\.env|secrets|credentials|\.aws|\.ssh|private_key)'
+    file_globs: ['.copilotignore', '.aiignore', '.cursorignore', '.continueignore', '.codeiumignore', '.vscode/settings.json', 'cursor.json', 'continue.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Sensitive paths excluded'
+    notes: 'Matched inside any AI-ignore file.'
+  - id: P16.3
+    control: DSGAI16
+    name: telemetry-off
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(telemetry[^\n]{0,10}(off|false|disabled)|share_data\s*[:=]\s*false)'
+    file_globs: ['.copilotignore', '.aiignore', '.cursorignore', '.continueignore', '.codeiumignore', '.vscode/settings.json', 'cursor.json', 'continue.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Telemetry off (PASS)'
+  - id: P16.4
+    control: DSGAI16
+    name: context-scope-limits
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(contextWindow|ignorePaths|excludeFiles|maxContextLines)'
+    file_globs: ['.copilotignore', '.aiignore', '.cursorignore', '.continueignore', '.codeiumignore', '.vscode/settings.json', 'cursor.json', 'continue.json']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Context scope limits'
+  - id: P17.1
+    control: DSGAI17
+    name: circuit-breaker
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(CircuitBreaker|@circuit|tenacity|resilience4j|pybreaker)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Circuit breaker (PASS)'
+    notes: 'LLM-calling module with none of P17.1-P17.5 = WARN.'
+  - id: P17.2
+    control: DSGAI17
+    name: retry-with-backoff
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(retry\s*\(|@retry|exponential_backoff|max_retries\s*=|backoff_factor\s*=|retry_with_backoff)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Retry with backoff (PASS)'
+  - id: P17.3
+    control: DSGAI17
+    name: timeout-on-llm-call
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(timeout\s*=\s*\d|request_timeout\s*=|httpx\.[^\n]{0,30}timeout)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Timeout on LLM call (PASS)'
+  - id: P17.4
+    control: DSGAI17
+    name: rate-limit-incoming
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(rate_limit|@throttle|RateLimiter|slowapi|flask_limiter|express-rate-limit)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Rate limit incoming (PASS)'
+  - id: P17.5
+    control: DSGAI17
+    name: fallback-response
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(fallback_response|default_response|graceful_degradation|on_failure_return)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Fallback response (PASS)'
+  - id: P17.6
+    control: DSGAI17
+    name: unbounded-retry
+    classification: structural
+    signal: fail
+    confidence: medium
+    pcre: 'while\s+True[^}]{0,200}(generate|complete|chat)|retry\s*\(\s*\)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Unbounded retry (FAIL)'
+  - id: P18.1
+    control: DSGAI18
+    name: output-guardrails
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(guardrails|nemo[._-]guardrails|llm_guard|output_guard|filter_output|sanitize_response|moderation)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Output guardrails (PASS)'
+  - id: P18.2
+    control: DSGAI18
+    name: pii-detection-on-output
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(detect_pii[^\n]{0,20}output|output[^\n]{0,20}detect_pii|presidio[^\n]{0,20}output|scan_output)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'PII detection on output'
+  - id: P18.3
+    control: DSGAI18
+    name: logprobs-exposed
+    classification: structural
+    signal: warn
+    confidence: low
+    pcre: '(logprobs\s*=\s*True|return_logprobs\s*=\s*True|include_logprobs\s*=\s*True|top_logprobs\s*=\s*\d)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Logprobs exposed (WARN)'
+  - id: P18.4
+    control: DSGAI18
+    name: llm-call-sites
+    classification: structural
+    signal: count
+    confidence: medium
+    pcre: '(\.chat\.completions\.create|client\.messages\.create|openai\.ChatCompletion|anthropic\.messages|generate_content|chat\.complete)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'LLM call sites (count)'
+    requires_nearby: {rule: P18.5, lines: 10}
+  - id: P18.5
+    control: DSGAI18
+    name: max-tokens-set
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: 'max_tokens\s*=\s*\d+'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'max_tokens set (PASS)'
+  - id: P19.1
+    control: DSGAI19
+    name: anonymization-before-export
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(anonymize_for_labeling|pseudonymize|redact_before_export|mask_for_labeling)'
+    file_globs: ['*.py', '*.ipynb', '*.ts', '*.yaml']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Anonymization before export'
+    gated_on: labeling
+  - id: P19.2
+    control: DSGAI19
+    name: labeling-sdk-auth-from-vault
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(label_studio[^\n]{0,30}vault|labelbox[^\n]{0,30}secret|scale_api_key[^\n]{0,30}(vault|env))'
+    file_globs: ['*.py', '*.ipynb', '*.ts', '*.yaml']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Labeling SDK auth from vault'
+    gated_on: labeling
+  - id: P19.3
+    control: DSGAI19
+    name: labeling-access-audit
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(label[._-]audit|labeling_access_log|annotator_audit)'
+    file_globs: ['*.py', '*.ipynb', '*.ts', '*.yaml']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Labeling access audit'
+    gated_on: labeling
+  - id: P19.4
+    control: DSGAI19
+    name: re-identification-post-label
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(reid_check_post_label|labeling_reidentification|post_label_validation)'
+    file_globs: ['*.py', '*.ipynb', '*.ts', '*.yaml']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Re-identification post-label'
+    gated_on: labeling
+  - id: P20.1
+    control: DSGAI20
+    name: api-auth-required
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(api_key[^\n]{0,15}(verify|validate)|bearer_token[^\n]{0,15}validate|Authorization[^\n]{0,15}required|@require_auth|auth_middleware|Depends\(.*auth)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'API auth required (PASS)'
+  - id: P20.2
+    control: DSGAI20
+    name: rate-limit
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(rate_limit|RateLimiter|@throttle|slowapi|flask_limiter|express-rate-limit|@limiter\.limit)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Rate limit (PASS)'
+  - id: P20.3
+    control: DSGAI20
+    name: input-length-validation
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(max_input_length\s*=|max_input_tokens\s*=|input[._-]truncat|validate[._-]length)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Input length validation (PASS)'
+  - id: P20.4
+    control: DSGAI20
+    name: prompt-injection-detect
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(prompt_injection_detect|detect_injection|llm_guard|rebuff|input_guard|nemo[._-]guard)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Prompt injection detect (PASS)'
+  - id: P20.5
+    control: DSGAI20
+    name: inference-endpoint
+    classification: structural
+    signal: count
+    confidence: medium
+    pcre: '(@app\.(post|get)\(["''][^"'']*(chat|generate|completion|infer|predict)|@router\.(post|get)\(["''][^"'']*(chat|generate|completion|infer|predict)|app\.(post|get)\(["''][^"'']*(chat|generate|completion|infer|predict))'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt', '*.go']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Inference endpoint (count)'
+    requires_nearby: {rules: ['P20.1', 'P20.2'], lines: 15}
+  - id: P21.1
+    control: DSGAI21
+    name: read-only-retrieval
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(read_only\s*=\s*True|readonly\s*=\s*True|HttpMethod\.GET|http_method[^\n]{0,10}get)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Read-only retrieval (PASS)'
+  - id: P21.2
+    control: DSGAI21
+    name: kb-write-from-agent
+    classification: structural
+    signal: warn
+    confidence: low
+    pcre: '(chromadb|qdrant|pinecone|weaviate|knowledge_base|kb_client)[^\n]{0,40}\.(upsert|insert|update|delete|add_documents|write)\('
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'KB write from agent (WARN)'
+    notes: 'P21.2 in an agent module without P21.3 = WARN.'
+  - id: P21.3
+    control: DSGAI21
+    name: write-content-validation
+    classification: structural
+    signal: info
+    confidence: low
+    pcre: '(validate_content|sanitize[._-]write|verify[._-]knowledge|content[._-]check[._-]write)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'Write content validation'
+  - id: P21.4
+    control: DSGAI21
+    name: kb-versioning
+    classification: structural
+    signal: pass_signal
+    confidence: medium
+    pcre: '(versioned\s*=\s*True|audit_writes|version_id|kb_audit|knowledge[._-]audit)'
+    file_globs: ['*.py', '*.ts', '*.java', '*.kt']
+    exclude_globs: []
+    framework: 'dsgai-2026-v1.0'
+    description: 'KB versioning (PASS)'
diff --git a/dsgai_scanner_tool/rules/rules.schema.json b/dsgai_scanner_tool/rules/rules.schema.json
new file mode 100644
index 0000000..5b392e8
--- /dev/null
+++ b/dsgai_scanner_tool/rules/rules.schema.json
@@ -0,0 +1,71 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://github.com/GenAI-Security-Project/GenAI-Data-Security-Initiative/dsgai_scanner_tool/rules/rules.schema.json",
+  "title": "DSGAI scanner ruleset",
+  "description": "Schema for rules/dsgai-rules.yaml. Note: JSON Schema cannot verify that a `pcre` string actually compiles under rg --pcre2 — the self-test CI (PR-06) does that.",
+  "type": "object",
+  "required": ["ruleset_version", "framework", "rules"],
+  "additionalProperties": false,
+  "properties": {
+    "ruleset_version": { "type": "string", "pattern": "^[0-9]+\\.[0-9]+\\.[0-9]+$" },
+    "framework": { "type": "string", "pattern": "^dsgai-[0-9]{4}-v[0-9]+\\.[0-9]+$" },
+    "rules": {
+      "type": "array",
+      "minItems": 1,
+      "items": { "$ref": "#/$defs/rule" }
+    }
+  },
+  "$defs": {
+    "ruleId": { "type": "string", "pattern": "^P[0-9]{2}\\.[0-9]+$" },
+    "rule": {
+      "type": "object",
+      "required": [
+        "id", "control", "name", "classification", "signal",
+        "confidence", "pcre", "file_globs", "framework", "description"
+      ],
+      "additionalProperties": false,
+      "properties": {
+        "id": { "$ref": "#/$defs/ruleId" },
+        "control": { "type": "string", "pattern": "^DSGAI[0-9]{2}$" },
+        "name": { "type": "string", "pattern": "^[a-z0-9]+(-[a-z0-9]+)*$" },
+        "classification": { "enum": ["structural", "value_bearing"] },
+        "signal": { "enum": ["fail", "warn", "pass_signal", "count", "info"] },
+        "confidence": { "enum": ["high", "medium", "low"] },
+        "pcre": { "type": "string", "minLength": 1 },
+        "file_globs": {
+          "type": "array",
+          "items": { "type": "string", "minLength": 1 }
+        },
+        "exclude_globs": {
+          "type": "array",
+          "items": { "type": "string", "minLength": 1 }
+        },
+        "framework": { "type": "string", "pattern": "^dsgai-[0-9]{4}-v[0-9]+\\.[0-9]+$" },
+        "description": { "type": "string", "minLength": 1 },
+        "remediation": { "type": "string" },
+        "references": { "type": "array", "items": { "type": "string" } },
+        "subtract": {
+          "type": "array",
+          "items": { "$ref": "#/$defs/ruleId" }
+        },
+        "requires_nearby": {
+          "type": "object",
+          "additionalProperties": false,
+          "properties": {
+            "rule": { "$ref": "#/$defs/ruleId" },
+            "rules": { "type": "array", "items": { "$ref": "#/$defs/ruleId" } },
+            "lines": { "type": "integer", "minimum": 0 },
+            "scope": { "enum": ["module", "file"] },
+            "absent": { "type": "boolean" }
+          },
+          "anyOf": [
+            { "required": ["rule"] },
+            { "required": ["rules"] }
+          ]
+        },
+        "gated_on": { "enum": ["multimodal", "synthetic_data", "labeling"] },
+        "notes": { "type": "string" }
+      }
+    }
+  }
+}