Skip to content

Inject persistent skills into agents (codex) - #2444

Merged
numnx merged 1 commit into
feature/CODUX-149-self-reflection-persisten-skillsfrom
task/feature-codux-149-self-reflection-persisten-skil-t04-codex-mr9zbb7w
Jul 7, 2026
Merged

Inject persistent skills into agents (codex)#2444
numnx merged 1 commit into
feature/CODUX-149-self-reflection-persisten-skillsfrom
task/feature-codux-149-self-reflection-persisten-skil-t04-codex-mr9zbb7w

Conversation

@numnx

@numnx numnx commented Jul 7, 2026

Copy link
Copy Markdown
Collaborator

🧩 Task: Inject persistent skills into agents

Automated task PR opened by Code UX for T04.

📋 Summary

Sprint: Sprint 149 · Self Reflection & Persisten Skills · Task: T04

Add disabled-by-default persistent skill storage for shareable agent self-enhancement and a configurable self-reflection improvement loop for planning and QA agents, with MCP management/search surfaces, runtime prompt injection, dashboard controls, tests, and docs.

📝 Full Task Prompt
## Objective
Make enabled persistent skill storages available to invoked agents through prompt instructions, safe storage mounts, and skill retrieval MCP access.

## Scope
- `src/services/cli-workflow-service.ts`
- `src/services/provider-execution-service.ts`
- `src/infrastructure/providers/cli/provider-runner.ts`
- `src/infrastructure/providers/cli/workspace-manager.ts`
- `src/services/structured-provider-response-service.ts`
- `src/services/skill-service.ts`
- `src/services/agent-mcp-access.ts`
- `src/repositories/agent-preset-repository.ts`
- `tests/backend/services/cli-workflow-service.test.ts` or nearest CLI workflow suite
- `tests/backend/infrastructure/providers/cli/workspace-manager.test.ts` or nearest provider runtime suite
- `docs/architecture/agent-preset-foundation.md`
- `docs/settings/provider-routing.md`

## Implementation Requirements
1. Resolve the invoked agent preset's attached persistent skill storages before CLI task, planning, QA, dashboard-reply, clarification, and remediation provider calls where an agent preset is known.
2. Append a clear disabled-by-default runtime instruction only when the agent has persistent skill storage enabled and at least one attached storage; the instruction must explain how to search existing skills and how to save a newly created skill through MCP or the mounted persistent path.
3. Mount persistent skill storage into Docker/host provider workflows at a path outside the project workspace, using project/agent-safe directory names and read/write permissions only for attached storages.
4. Ensure provider MCP configuration exposes the skill retrieval surface from T03 to eligible agents without automatically broadening unrelated Code UX management tools.
5. Keep existing memory context injection and `workerLearningsInstruction` behavior unchanged; persistent skills should be an additional section, not a replacement for memory.
6. Add tests proving disabled agents get no skill instruction or mount, enabled agents receive attached storage paths and search guidance, and workspace paths are never used as the persistent storage root.
7. Update runtime/provider docs to describe storage isolation, mount paths, and how agents should use search before creating duplicate skills.

## Constraints
- Depends on T02 for service resolution and T03 for retrieval MCP behavior.
- Do not enable persistent storage by default for existing or new agents.
- Do not mount arbitrary user-provided paths from settings; derive controlled paths under the Code UX application data directory.
- Do not expose real project names in docs, tests, or log snapshots.

## Verification
- Run `pnpm run lint`.
- Run focused backend provider/runtime tests for CLI workflow and workspace manager.
- Run `pnpm run test:backend` if focused runtime suites are not available.
- Success means persistent skill storage is isolated from workspaces, opt-in only, mounted/instructed for attached agents, and absent from disabled agents.

✅ QA Review

⏳ QA review runs after this PR is opened. Check the task card in the dashboard for the outcome once it lands.

🤖 Model & Provider

Provider Codex CLI
Model gpt-5.5

⏱️ Timing

Started 2026-07-07 01:37:20 UTC
Finished 2026-07-07 01:51:31 UTC
Duration 14m 10s

📊 Token Usage

Metric Value
Input tokens 341,960
Cached input tokens 11,580,800
Output tokens 25,206
Total tokens 11,947,966
Invocations 1
Tool calls 132
Active time 13m 40s

Included usage estimate (subscription/local login): $8.26

🌿 Branch Info

Base: feature/CODUX-149-self-reflection-persisten-skills
Head: task/feature-codux-149-self-reflection-persisten-skil-t04-codex-mr9zbb7w


🤖 Generated by Code UX

@numnx
numnx merged commit a05ac1d into feature/CODUX-149-self-reflection-persisten-skills Jul 7, 2026
6 checks passed
@numnx
numnx deleted the task/feature-codux-149-self-reflection-persisten-skil-t04-codex-mr9zbb7w branch July 7, 2026 01:53
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant