Skip to content

feat(code): add agent-qa workflow to shared QA skills - #344

Open
pranshuchittora wants to merge 1 commit into
closedloop-ai:mainfrom
pranshuchittora:feat/agent-qa-skill
Open

pranshuchittora wants to merge 1 commit into
closedloop-ai:mainfrom
pranshuchittora:feat/agent-qa-skill

Conversation

@pranshuchittora

Copy link
Copy Markdown

Adds closedloop-core:agent-qa for selected web/mobile regressions, saved-run evidence review and memidex provenance. agent-qa from vostride combines natural-language tests with advanced memory and self-improvement. The skill uses an already configured external CLI, preserves recorded verdicts and build scope, and keeps read-only review separate from execution or memory changes.

The original instructions are Apache-2.0; the separately licensed agent-qa runtime is not bundled. Both core manifests advance from 1.5.2 to 1.6.0. The repository's /update-documentation workflow updated the changelog and plugin README, including its required independent inventory review.

The required version bump exposed a hardcoded launcher-test replacement version (1.5.3). The fixture now derives a newer patch version from the manifest, preserving the cache-update and fallback assertions; runtime launcher behavior is unchanged.

Validation:

  • uv sync --frozen --group dev with Python 3.13.
  • uv run --frozen pytest plugins/closedloop-core/tools/python/ -q — 90 passed, including the new skill in the existing frontmatter contract.
  • uv run --frozen ruff check . and uv run --frozen pyright — pass.
  • In tools/plugin-manifests: npm ci, npm run typecheck, npm test — six tests pass.
  • claude plugin validate plugins/closedloop-core, skill frontmatter, five shell examples and git diff --check — pass.
  • skills 1.7.4 local installation for Claude Code and Codex — both skill files match their sources in each installed copy.

Validation covers the contributed instructions, packaging and affected plugin. No live agent-qa application/provider run or harness compatibility evaluation is claimed.

- Add selected-test and saved-run guidance to closedloop-core.
- Advance both manifests and generate documentation for 1.6.0.
- Derive the launcher test replacement version from the manifest.

Testing: 90 core tests, six manifest tests, lint and types pass.
Local installer copies match; plugin and skill validation pass.

Risks: External agent-qa execution still needs a configured runtime.
No live application or provider execution was performed.
@shafty023

Copy link
Copy Markdown
Collaborator

Beginning a code-review. Please allow time for me to finish before merging the PR

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.

2 participants