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

qa-report

Skill2 starsChanged 7 days agoApache-2.0View on GitHub

Aggregate the evidence directory into QA metrics — pass rate, P0/P1 trend, BLOCKED reasons, routing quality, learning throughput, per project and product. Use for "/qa-report", weekly/sprint summary, "QA-Bericht", "Testreport", "wie läuft's mit dem Testing". Not for a single ticket's verdict — /report-to-jira.

8 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 0

    Distinctive

    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:

    Every run writes execution_summary.json.

    line 9
  2. Very specific:

    Run the aggregator (it reads only evidence/, writes nothing):

    line 16
  3. Very specific:

    Read the output and interpret it against the project's own gates — the defaults in .claude/rules/core/qa-principles.md are overridden per …

    line 24

“Common” is measured: kept by at least 6 in 10 of the 7 edited versions of this file we could read. “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

/qa-report — what the runs already know

Every run writes execution_summary.json; nothing ever read one back. This skill turns that pile into the numbers a QA lead, a release manager or a retro actually asks for. No new data source, no MCP server, no process change — just the evidence that is already there.

Steps

  1. Run the aggregator (it reads only evidence/, writes nothing):

    python3 .claude/skills/qa-report/aggregate.py --since 30
    # --project SHOP    one project only
    # --json            machine-readable, for a dashboard or a pipeline
  2. Read the output and interpret it against the project's own gates — the defaults in .claude/rules/core/qa-principles.md are overridden per project in PROJECT.md → "Pipelines / quality gates", and a product may tighten them further. A pass rate of 88 % is a breach in a project that declared ≥ 90 % and fine in one that declared ≥ 80 %. Never report a number without the gate it is measured against.
  3. Turn the two blueprint-specific columns into actions, because they are the ones nobody else will read:
    • Routing needed a human — every user, default or secondary resolution is a missing component/label row in the project's registry. Propose the rows; retro-learn promotes them (.claude/rules/core/product-layer.md).
    • Awaiting ruling — every test-outdated or test-brittle failure still waiting on a human is a test the suite is currently lying about. One or two is normal traffic; a number that grows week over week means the suite is drifting away from the product, and the fix is a grooming session with the test owners, not a metric. Report the count and the age of the oldest, never a percentage.
    • Undeclared learnings — a run that produced evidence without declaring its learnings means the evidence gate did not fire. That is an enforcement gap, not a reporting detail: check that the gate is wired in this repo before trusting any other number in the report.
  4. Deliver it where the audience is. A QA lead does not read ticket comments:
    • a published page for the team (offer it — the report is Markdown already),
    • a Confluence page update if the project keeps one (per PROJECT.md → "Confluence"; note that editing Confluence is an explicit, separate request, not something this skill does on its own),
    • or a JIRA comment on the project's regression ticket, per report-to-jira.
  5. Anything the report reveals about the blueprint itself (a metric that is always empty, a scope that never appears) is a retro-learn capture.

What this skill does not do

Invent metrics the schema cannot support. The aggregator prints an explicit "not computable" section — defect escape rate and coverage are not derivable from execution_summary.json, and the flake rate covers only runs that reported test_cases[] (the CLI runner; never an exploratory session). Estimating the rest anyway would produce exactly the kind of sourced-looking-but-wrong number this blueprint keeps warning about. The flake rate is the worked example of the right order: the schema grew attempts first, then the aggregator started counting.

It also does not judge people. The unit of every number here is a run, not a tester.

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.