Add spam risk reviewer skill#146
Open
Li-AmG wants to merge 79 commits into
Open
Conversation
default to the /internal/thread-outbox pending cursor (drops the fragile client cursor-cache that re-walked history on a fresh runner) and listen for a board-sync repository_dispatch so the venue can trigger the drain low-latency.
a no-ua, no-browser-headers, http1.1 client is an obvious bot signature. the fetch tool now presents a current chrome ua + the browser header set and negotiates http2 with gzip/brotli, applied as overridable defaults. configurable via RUNX_HTTP_USER_AGENT and RUNX_HTTP_BROWSER=0; the anthropic and registry transports stay plain; all transport guards unchanged. tls (ja3/ja4) and http2 fingerprint matching are out of scope.
Conventional follow-up for the pushed lockfile refresh.
Adds the dependency-cve-audit runx skill and registers it in the official catalog. Verified: - CI green on PR runxhq#82 - node --check skills/dependency-cve-audit/run.mjs - runx doctor skills/dependency-cve-audit --json - runx harness skills/dependency-cve-audit --receipt-dir <tmp> --json - packages/cli/src/skill-refs.test.ts
Adds the structured-extraction runx skill and completes the paid follow-up integration work. Maintainer cleanup added: - deterministic tool fixture for structured.extract - SKILL.md frontmatter for official catalog generation - official skill lock/Rust table/catalog allowlist entries Verified: - CI green on PR runxhq#80 - local merge simulation after runxhq#82 - node --check skills/structured-extraction/tools/structured/extract/run.mjs - runx doctor skills/structured-extraction --json - runx harness skills/structured-extraction --receipt-dir <tmp> --json - runx dev tools/structured/extract --json with RUNX_PROJECT_DIR set to the skill root - packages/cli/src/skill-refs.test.ts
Remove the user-facing installation-id flag from add/registry flows, keep native command help aligned, and update registry fixtures/docs to use versioned runx add plus runx skill execution commands.
Rename the bundled runx operator skill to ops-desk, remove product-specific fixture names, keep newer maturing skills internal until they meet the public catalog bar, and make graph skills fail closed when required graph inputs are missing.
Update the native CLI skill execution/export surfaces, add governed data-plane contracts and fixtures, refresh official skill catalog coverage, and remove local .ai state from Git tracking. Validation: pnpm bindings:check; pnpm exec tsc --noEmit --allowJs --checkJs --module NodeNext --moduleResolution NodeNext --target ES2022 --skipLibCheck scripts/check-upstream-skill-bindings.mjs; git diff --check
Replace the frantic-specific event-replay driver with a generic, stateless reconcile engine: consume a tenant's desired thread state, diff against live GitHub, apply only the difference via the provider primitives. Drop the cursor and the frantic_thread_outbox translator; generalize label ensure.
Reconcile read each issue thread once per run (state, labels, comment markers) and write only actual differences, instead of re-fetching the whole thread per comment. Steady state is now one read per thread and zero writes; cut redundant per-run observations to newly-created threads only.
… param, share verify path resolver
Signed-off-by: Li-AmG <288224810+Li-AmG@users.noreply.github.com>
Li-AmG
force-pushed
the
li-amg/spam-risk-reviewer-62
branch
from
c56eac7 to
e53329d
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this does
Adds a read-only
spam-risk-reviewerrunx skill for campaign send-risk preflight.How to use it
Run the local harness:
The skill accepts
campaign_draft,list_metadata, andsender_auth_posture, then emitssend_risk_verdict,evidence_json, andreport_md.How it works
The runner compares sender authentication and list hygiene against bounded policy thresholds. Clear sends return
risk_level: pass; authentication or list hygiene failures returnrisk_level: hold,preflight_clear: false, blockers, andneeds_humanescalation for send-as approval.Testing
Manual node dogfood completed for both harness fixtures:
low-risk-verified-sender->risk_level: pass,preflight_clear: truehigh-risk-incomplete-auth-poor-list->risk_level: hold,preflight_clear: false,needs_humanLocal runx harness could not be executed on this Windows host because the 0.6.13 native
runx.exepackage was blocked by the OS withEPERM/missing executable after install. The package files and fixtures are included for hosted harness review.