ci: add CI job to run post-script unit tests

[fullsend-ai-retro]

What happened

PR #29 added seven post-script test files totaling 3,086 lines:

• scripts/post-code-test.sh (788 lines)
• scripts/post-review-test.sh (683 lines)
• scripts/post-triage-test.sh (455 lines)
• scripts/validate-output-schema-test.sh (380 lines)
• scripts/post-prioritize-test.sh (327 lines)
• scripts/post-retro-test.sh (286 lines)
• scripts/post-fix-test.sh (167 lines)

These tests exercise critical post-script logic (PR title rewriting, comment posting, schema validation, error handling) using mock gh commands. However, no CI job runs them. The only CI workflow (lint.yml) has a test job that runs pre-commit run --all-files — it does not execute any of these test scripts.

This means a regression in any post-script (e.g., a broken regex in PR title rewriting, a missing error guard in schema validation) would not be caught until it fails during a live agent run.

What could go better

These tests represent significant investment in correctness verification for the most security-sensitive parts of the agent pipeline (the post-scripts handle GitHub API mutations, token usage, and content sanitization). Not running them in CI undermines that investment.

Confidence: High. The test scripts exist, they are self-contained (use mocks, no network access needed), and there is no CI job that invokes them. This is verifiable from the workflow file and the test script headers (each includes a comment like "Run from the repo root: bash scripts/post-retro-test.sh").

Proposed change

Add a new job to .github/workflows/lint.yml (or create a dedicated test.yml workflow) that runs all post-script tests:

script-tests:
  runs-on: ubuntu-24.04
  steps:
    - uses: actions/checkout@<sha>  # SHA-pinned
    - name: Run post-script tests
      run: |
        for test in scripts/*-test.sh; do
          echo "::group::${test}"
          bash "${test}"
          echo "::endgroup::"
        done

The tests require bash, jq, and standard Unix tools — all available on ubuntu-24.04 without additional setup. Some tests may also need python3 for JSON schema validation (check validate-output-schema-test.sh dependencies).

Validation criteria

After the change:

1. All seven test scripts pass in the new CI job on the current main branch
2. Introducing a deliberate regression in any post-script (e.g., removing a required field check in post-retro.sh) causes the corresponding test to fail in CI
3. The CI job completes within a reasonable time (these are unit tests with mocks, expected under 60 seconds total)

---

Generated by retro agent from #29

[fullsend-ai-triage]

🤖 Finished Triage · ✅ Success · Started 10:36 PM UTC · Completed 10:39 PM UTC
Commit: e8381e3 · View workflow run →

[fullsend-ai-triage]

Triage Summary

This issue is already addressed by PR #37 (ci: add Makefile and workflow for agent script tests), which is open and under review.

What PR #37 delivers

• A Makefile with script-test and test targets that run all seven scripts/*-test.sh suites with per-suite timing
• A .github/workflows/script-test.yml workflow triggered on push, pull_request, and merge_group for the main branch
• jsonschema==4.23.0 installed via uv for schema validation tests
• All actions SHA-pinned, minimal permissions (contents: read)

Current status of PR #37

• Both automated reviewers (fullsend-ai-review and qodo-code-review) have reviewed it
• Review findings addressed: jsonschema version pinned, false-positive .ONESHELL finding dismissed with local + CI verification
• Awaiting human approval — the PR modifies .github/workflows/ (a protected path requiring explicit human review)

Recommendation

Merge PR #37 to resolve this issue. No additional implementation work is needed. After merge, consider the two low-severity doc suggestions from the review: adding script-test.yml to the README workflows table and adding a brief "Testing" section documenting make test.

Blocked by:

• ci: add Makefile and workflow for agent script tests #37