oceanbase · AlexStocks · Aug 25, 2026 · Aug 21, 2026 · Aug 21, 2026 · Aug 21, 2026
diff --git a/docs/en/docs/reference/configuration.md b/docs/en/docs/reference/configuration.md
@@ -1,6 +1,6 @@
 ---
 title: Configuration
-description: PowerContext paths, Server, Client, inference, and Codex environment variables.
+description: PowerContext paths, Server, Client, inference, and Agent integration environment variables.
 ---
 
 # Configuration
@@ -52,7 +52,7 @@ Server settings use the `POWERCONTEXT_SERVER_` prefix.
 | `POWERCONTEXT_SERVER_INFERENCE_GENERATION_TIMEOUT_SECONDS` | `30` | Generation timeout |
 | `POWERCONTEXT_SERVER_INFERENCE_EMBEDDING_BATCH_SIZE` | `10` | Maximum texts sent in one embedding request |
 | `POWERCONTEXT_SERVER_RUNTIME_EXPERIENCE_SCHEDULE_SECONDS` | unset | Experience incubation interval; unset disables that job |
-| `POWERCONTEXT_SERVER_EXTERNAL_SKILLS` | unset | JSON object containing the host identity and explicit Codex Skill roots |
+| `POWERCONTEXT_SERVER_EXTERNAL_SKILLS` | unset | JSON object containing the host identity and explicit Agent Skill targets |
 
 Static bearer authentication is disabled by default. When enabled, API and MCP requests must include
 `Authorization: Bearer <token>`; the liveness and readiness endpoints remain public. Plain HTTP should remain on a
@@ -105,35 +105,54 @@ The same configured generation model gates explicit Experience generation, manag
 and external Skill import or fork. Without it, these operations return a capability error before persisting a
 Candidate. Candidate Review, exact reads, and external Skill scan/list/resolve continue to work.
 
-Experience incubation is a separate APScheduler job with its own persisted Source cursor. It requires both
-`POWERCONTEXT_SERVER_RUNTIME_EXPERIENCE_SCHEDULE_SECONDS` and
-`POWERCONTEXT_SERVER_INFERENCE_GENERATION_MODEL`. Each activation inspects at most 32 Sources and exposes only Content
-Sources whose metadata contains `"kind": "task-outcome"`. The Memory and Experience jobs share the scheduler sidecar
-under `POWERCONTEXT_HOME`, but keep independent job identities and business cursors. Unsetting one interval removes
-only that job. See [Create and review an Experience](../how-to/create-and-review-experience.md) for setup and
-verification steps.
+Experience incubation is a separate APScheduler job with its own persisted Source cursor. Enable it with:
 
-### External Codex Skills
+```bash
+export POWERCONTEXT_SERVER_RUNTIME_EXPERIENCE_SCHEDULE_SECONDS=30
+export POWERCONTEXT_SERVER_INFERENCE_GENERATION_MODEL=provider:model-name
+powercontext server run
+```
 
-Configure host-local roots as one JSON value:
+Each activation inspects a fixed window of at most 32 Sources and exposes only Content Sources whose metadata contains
+`"kind": "task-outcome"` to the model. It creates pending Experience Candidates in the Review Inbox; it does not
+approve them, place them in PreparedContext, create a managed Skill, export it to an Agent target, or execute anything.
+The Memory and Experience jobs share the APScheduler sidecar under `POWERCONTEXT_HOME`, but keep independent job
+identities and business cursors. Unsetting one interval removes only that job.
+See [Create and review an Experience](../how-to/create-and-review-experience.md) for setup and verification steps.
+
+### Agent Skill targets
+
+Configure Codex and Claude Code host-local targets as one JSON value:
 
 ```bash
 export POWERCONTEXT_SERVER_EXTERNAL_SKILLS='{
   "host_id": "workstation-1",
-  "codex_roots": [
+  "targets": [
+    {
+      "target_id": "codex-project",
+      "agent_kind": "codex",
+      "installation_scope": "project",
+      "path": "/srv/project/.agents/skills",
+      "allow_managed_publish": true
+    },
     {
-      "root_id": "repository",
+      "target_id": "claude-project",
+      "agent_kind": "claude_code",
       "installation_scope": "project",
-      "path": "/srv/project/.agents/skills"
+      "path": "/srv/project/.claude/skills",
+      "allow_managed_publish": true
     }
   ]
 }'
 ```
 
-Root IDs must be unique. Supported installation scopes are `user`, `project`, and `plugin`. PowerContext scans only
-the immediate Skill package directories under these explicit roots; it does not infer a home directory, install
-packages, or grant execution authority. The `host_id`, locator, and registration are local-environment state, not a
-cross-host or cross-Agent contract.
+Target IDs must be unique. `agent_kind` supports `codex` and `claude_code`; installation scopes are `user`, `project`,
+and `plugin`. PowerContext scans only the immediate Skill package directories under these explicit targets; it does not
+infer a home directory, install packages, or grant execution authority. `allow_managed_publish` defaults to `false`;
+when true, the authenticated Skills Library or Review page may explicitly create or safely update an approved managed
+Skill in that target. The page still cannot submit an arbitrary path or overwrite a foreign or modified package. The
+`host_id`, locator, and registration are local-environment state, not a cross-host contract. Existing `codex_roots`
+configuration remains accepted as a Codex-only compatibility form; new configuration should use `targets`.
 
 The Server always creates non-recording OpenTelemetry request context so `X-PowerContext-Request-ID` can be derived from the
 inbound span. To enable recording and export for a CLI-managed Server, install

diff --git a/docs/en/docs/reference/interfaces.md b/docs/en/docs/reference/interfaces.md
@@ -1,6 +1,6 @@
 ---
 title: Interfaces
-description: Choose between the Codex plugin, DeepSeek Harness plugin, Pi package, CLI, Python SDKs, HTTP, and MCP.
+description: Choose between the Codex and Claude Code plugins, DeepSeek Harness plugin, Pi package, CLI, Python SDKs, HTTP, and MCP.
 ---
 
 # Interfaces
@@ -202,21 +202,42 @@ For the relationship between evidence, Candidate versions, approved Revisions, r
 
 ## Scheduled Experience incubation
 
-The scheduler accepts only Content Sources with metadata `"kind": "task-outcome"`, uses an independent persisted
-cursor, and creates pending Experience Candidates. Candidate writes and cursor advancement commit together. The job
-does not approve content or include pending content in `PreparedContext`. Setup and verification steps are in
+An integration can capture a completed task as a Content Source with metadata `"kind": "task-outcome"`. When the
+Experience schedule is configured, APScheduler scans bounded Source windows and asks the configured schema-bound
+pipeline for reusable situation, action, outcome, and lesson proposals. Each proposal cites exact Sources and enters
+the Review Inbox as a pending Experience Candidate.
+
+Experience incubation has its own persisted Source cursor, independent from Memory extraction. Candidate writes and
+cursor advancement commit together; a generation or write failure leaves the window available for retry. Ordinary
+prompt Sources are not Task Outcomes and are ignored by this job.
+
+Scheduling stops at the review boundary. It never approves an Experience, includes pending content in
+PreparedContext, derives a managed Skill, exports a Skill for an Agent target, or executes instructions. Skill authoring and
+export remain explicit steps after the supporting Experience is approved.
+Setup and verification steps are in
 [Create and review an Experience](../how-to/create-and-review-experience.md).
 
-## Managed Skill export to Codex
+## Managed Skill export to Agent targets
+
+A configured generator can produce complete managed Skill content through `generate_skill`; a human or integration
+can submit already-complete typed content through `propose_skill`. The proposal contains a name, discovery
+description, instructions, validation checks, and exact Source or Artifact lineage. It remains a Candidate until a
+reviewer approves the exact Candidate version.
+
+Approval creates an immutable Skill Revision. It does not install the Skill or grant execution authority. To make one
+approved Revision available to Codex or Claude Code, export it explicitly into a configured repository, user, or plugin
+Skill target. The projection writes `SKILL.md` and `powercontext.json`; the manifest records the Agent kind, exact
+Artifact reference and rendered-content hash. It refuses to replace an existing destination, so updates require an
+intentional new export rather than a silent overwrite.
 
-Approval creates an immutable Skill Revision but does not install it or grant execution authority. The Codex exporter
-writes `SKILL.md` and `powercontext.json` into a new destination and refuses to replace an existing directory. The
-manifest binds the projection to one exact Artifact Revision and rendered-content hash. See
-[Create and export a managed Skill](../how-to/create-and-export-skill.md) for the procedure.
+Codex can discover a repository-local export under `.agents/skills/<name>/SKILL.md`. The Artifact Revision remains
+the content authority; Claude Code uses `.claude/skills/<name>/SKILL.md` for the equivalent project target. Both
+directories are host-local projections that can be rebuilt from the same exact Revision.
+See [Create and export a managed Skill](../how-to/create-and-export-skill.md) for the procedure.
 
 ## External Agent-native Skills
 
-External Skills remain authoritative in their original local packages. With explicitly configured Codex roots, the
+External Skills remain authoritative in their original local packages. With explicitly configured Agent targets, the
 Server can scan a scope-local, rebuildable Registry and report name, description, provider, Agent kind, host,
 installation scope, locator, and whole-package fingerprint. Exact resolve succeeds only when the same package remains
 readable on the configured host and its fingerprint still matches. It never installs a package or falls back to a
@@ -234,8 +255,8 @@ managed Artifact.
 | --- | --- | --- | --- | --- |
 | External Agent-native Skill | Original package | No for scan/list/resolve; yes for import/fork | No for discovery; yes after import/fork | Host-local Registry and exact resolve |
 | Experience | Exact approved Artifact Revision | Yes for generate/evolve; no for typed `propose` | Yes | Exact read and approved-head FTS recall in PreparedContext |
-| Managed Skill | Exact approved Artifact Revision | Yes for generate/evolve/import/fork; no for typed `propose` | Yes | Exact read and explicit Codex projection |
-| Codex projection | Its source managed Skill Revision | No | No additional review | Rebuildable host-local copy |
+| Managed Skill | Exact approved Artifact Revision | Yes for generate/evolve/import/fork; no for typed `propose` | Yes | Exact read and explicit Agent projection |
+| Agent projection | Its source managed Skill Revision | No | No additional review | Rebuildable Codex or Claude Code host-local copy |
 
 ## Core SDK