♻️ refactor(agent-signal): S6 — migrate skillManagement to execAgent builtin agent#15443
Conversation
…builtin agent Move the self-iteration skill-management action off the inline policy implementation onto an execAgent-dispatched builtin agent (slug `skill-management`), mirroring the S3/S4 memoryWriter + self-iteration migration. Adds the `agentSignalSkillManagement` serverRuntime, the builtin-tool-agent-signal skill-management manifest/systemRole, and the builtin-agents skill-management agent; strips the ~3.5k-line inline skillManagement policy down to the dispatch shim. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 75cc790798
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| marker, | ||
| prompt, | ||
| slug: BUILTIN_AGENT_SLUGS.skillManagement, | ||
| ...(messageId ? { sourceMessageId: messageId } : {}), |
There was a problem hiding this comment.
Anchor skill runs to the assistant message
When this action comes from the hydrated client.runtime.complete source, action.payload.messageId is the parent user message; the completed assistant message is encoded in action.source.sourceId as <assistantId>:completion:<parentId> (the memory planner extracts that for this exact thread-anchoring case). Passing the user message here means the new isolated skill-management thread, and the marker/receipt anchor built from the same value just above, attach to the wrong chat message or fail to show where users expect the post-turn skill update.
Useful? React with 👍 / 👎.
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## canary #15443 +/- ##
==========================================
- Coverage 70.88% 70.83% -0.05%
==========================================
Files 3222 3223 +1
Lines 319610 318399 -1211
Branches 28126 33610 +5484
==========================================
- Hits 226557 225550 -1007
+ Misses 92877 92673 -204
Partials 176 176
Flags with carried forward coverage won't be shown. Click here to find out more.
🚀 New features to boost your workflow:
|
💻 Change Type
🔗 Related Issue
Related to LOBE-9434. Follow-up to #15392 (S1–S4 execAgent migration, already merged).
🔀 Description of Change
S6 of the agent-signal execAgent migration. Moves the self-iteration skill-management action off its inline policy implementation onto an
execAgent-dispatched builtin agent (slugskill-management), mirroring the S3/S4 memoryWriter + self-iteration pattern.agentSignalSkillManagementserverRuntime (src/server/services/toolExecution/serverRuntimes/) and register it in itsindex.ts.skill-managementmanifest + systemRole, and wire the new tool surface throughbuiltin-toolsidentifiers.ts+index.ts(discoverable:false,hidden:true) so the schema is generated.skill-managementbuiltin agent inbuiltin-agents(slug registered inBUILTIN_AGENT_SLUGS/BUILTIN_AGENTS/SELF_ITERATION_AGENT_SLUGS).skillManagement.tspolicy down to the dispatch shim; project the completion-side receipt viabuildSelfIterationReceiptsand remove the sync projection fromreceiptService.Net: 22 files changed, ~359 insertions, ~3544 deletions.
🧪 How to Test
skillManagement.test.tsandreceiptService.test.tspass. End-to-end self-iteration skill runs verified locally vialh agent-signal trigger(BDDskill-writercase in the cloud repo).📝 Additional Information
Two registrations are required for a new self-iteration tool surface (serverRuntime + builtin-tools manifest); both are included here so the builtin agent uses the
agent-signal-*tools rather than falling back to default tools.🤖 Generated with Claude Code