agentleFS
Sn4kyGitSn4kyGit/qa-agent-blueprint/.claude/skills/retest-bug/SKILL.md

retest-bug

Skill2 starsChanged 7 days agoApache-2.0View on GitHub

Focused re-test of a fixed bug plus a tight regression sweep — scoped from the product's flow-map.md Neighbors when available, risk-based otherwise. Use for "/retest-bug <KEY-123>", "re-test <KEY-123>", "teste <KEY-123> nochmal", "gegentesten", or when a fix has landed on stage ("der Fix ist auf Stage", "fix is deployed/verified on stage — retest it"). Not for first-time tests — test-ticket.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run preflight.sh <KEY> before any browser/JIRA call.

    line 16
  2. Very specific:

    Read the original bug report (description, expected-vs-actual and exact repro, plus the last relevant comments) and route once to …

    line 19
  3. Distinctive:

    Read the effective Execution mode and scope the work:

    line 34
  4. Distinctive:

    Do not soften this into a "partial pass." Report FAIL, keep the original ticket open/reopened per the project's FAIL transition, and note …

    line 86
  5. Very specific:

    After classification as a product defect, file it as a new bug via bug-report (do not conflate it with the original ticket) and …

    line 94

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/retest-bug — verify a fix + surrounding regression

A narrower variant of test-ticket for bugs with a known repro. It uses the same dispatcher: the agent loading this skill is the main orchestrator and performs preflight, intake, routing and readiness once; only a complex re-test hands the remaining orchestration to qa-manager.

Steps

  1. Run preflight.sh <KEY> before any browser/JIRA call. Do not navigate to stage for liveness yet. JIRA access browser may use the playwright-driven JIRA UI for intake now; that exception never includes stage.
  2. Read the original bug report (description, expected-vs-actual and exact repro, plus the last relevant comments) and route once to projects/<KEY>/PROJECT.md + learnings.md, then to a product only per the ## Products registry. Write intake.json with the repro, expected build, risk and regression area. If the repro is vague/incomplete, extract only ticket-held environment, build, steps and data now; mark live bug-reproduction as a Complex-path escalation after readiness, pointed at the routed .agents/qa-project-context/<KEY>[/PRODUCT].md bridge. Do not reproduce on stage during intake.
  3. Apply docs/pipeline/01a-readiness.md and write readiness.json before any stage navigation, recon or spec run: project-allowed status, no blockers, linked MR/PR merged when required/present under the merge policy, expected fix build deployed when verification is required. Prefer API/CLI. If the build is missing, use deploy-to-stage only when allowed and then re-verify; otherwise BLOCKED. Only a ready result permits the one stage-liveness navigation and execution.
  4. Read the effective Execution mode and scope the work:
    • cli-first / hybrid — an existing spec covering the repro and its tight regression sweep runs through test-automator. New specs still require recon and the red check; hybrid's configured live verification pass still applies.
    • mcp-only, or an escalation trigger (non-reproducible bug, visual fix, non-assertable behavior) — live through test-executor. Never escalate merely because a test is red.
  5. Seal evidence/<KEY>/<TICKET>/<RUN>/context-capsule.json once per docs/pipeline/context-capsule.md, with exact intake/readiness, effective config, relevant confirmed knowledge and source hashes. Run context-capsule.py seal and one full validate using --qa-home and --run-dir before dispatch. Revision 1 is context-capsule.json; revision N>1 must be context-capsule.vN.json in the same run dir with every predecessor preserved. Pass only the selected revision path in Task prompts. A changed build/readiness or effective context returns to this orchestrator for readiness re-verification and a new immutable revision. Changed same/older-revision payloads conflict; identical reseals are idempotent and invalid existing hashes are never overwritten.
  6. Dispatch by complexity:
    • Fast Path — unambiguously testable, low risk, no escalation, exactly one mode-conformant execution branch: the main orchestrator invokes test-executor-cheap OR test-automator-cheap directly with the selected sealed revision path and repro/sweep charter. No PROJECT/PRODUCT full reads, re-routing, mode/suite lookup or repeated Learnings merge in the leaf. No qa-manager.
    • Complex path — invoke qa-manager with the selected sealed revision path, run dir and any recon/plan. It consumes the capsule without reading the ticket or PROJECT/PRODUCT files, routing, repeating readiness, or redoing mode/suite lookup and Learnings merging, and owns the remaining phases.
  7. Execute the exact original repro first, then 2–4 adjacent checks, not the whole application. Prefer the fixed flow's Neighbors from confirmed Areas in the routed flow-map.md and record that provenance. Unconfirmed candidates, basis-less Neighbors and stale or absent maps are not binding scope: use risk-based judgment around the actual defect instead. Keep the sweep in the same mode, with live observation only where a real escalation trigger requires it. When a pre-fix build/commit is available, proving the old failure there and the fix on the expected build is stronger evidence; never deploy or roll back solely to manufacture it.
  8. On FAIL, the owning orchestrator invokes failure-analyst with the sealed selected revision path and failure packet — main on the Fast Path, qa-manager on the complex path; never the execution leaf. Distinguish a genuine product defect from a missing/wrong build (environment → BLOCKED), a flake, or an original repro made outdated by the deliberate fix. Only product-defect and undetermined justify reopening; class-specific evidence and reporting stay unchanged.
  9. Produce execution_summary.json and all mandatory evidence, then use report-to-jira on PASS and the classified FAIL/BLOCKED/NEEDS_REVIEW paths as in test-ticket. The summary's skills_used names every skill this re-test actually invoked (this one, recon, failure-analysis, …) — never merely planned ones (.claude/rules/core/evidence-policy.md). Use bug-reporter only after the required classification; shadow mode and transitions are unchanged. Close with the learnings declaration.

If the bug is not actually fixed

Do not soften this into a "partial pass." Report FAIL, keep the original ticket open/reopened per the project's FAIL transition, and note precisely what part of the original repro still reproduces. Do not report "not fixed" when the truth is "not deployed" or the recorded repro no longer matches the deliberately changed flow.

If a regression was introduced

After classification as a product defect, file it as a new bug via bug-report (do not conflate it with the original ticket) and cross-reference both tickets.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.