agentleFS
seancabalseseancabalse/agentic-playbook/skills/verify-feature/SKILL.md

verify-feature

Skill0 starsChanged 15 days agoMITView on GitHub

Verify a feature in the running product against its acceptance criteria and states, with evidence. Use before you call a feature done.

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

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

  • 0

    Common

    Most of the 8 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Start only what's needed (the local backend, the app in a browser, the service).

    line 15
  2. Distinctive:

    Use the project's runbook for launch commands and the test accounts (a scripted second user for multi-user flows).

    line 16
  3. Very specific:

    Do the impact assessment in ${CLAUDE_PLUGIN_ROOT}/principles/08-testing-qa.md, "Tests for a change".

    line 21
  4. Distinctive:

    Run only the automated tests that the change can affect.

    line 23
  5. Distinctive:

    Put the impact assessment in your report.

    line 26
  6. Distinctive:

    Data: check the rows, files and jobs the action should create, with direct queries, counts and outcomes.

    line 34
  7. Very specific:

    When another agent or session continues from here, write a handoff with these results.

    line 47

“Common” is measured: kept by at least 6 in 10 of the 8 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

Verify in the real app

Playbook files named here: when the project has its own copy at .playbook/local/<same path>, use that copy instead.

Target: $ARGUMENTS

Setup

  • Start only what's needed (the local backend, the app in a browser, the service). Note each process so you can stop it.
  • Use the project's runbook for launch commands and the test accounts (a scripted second user for multi-user flows).
  • Pick the driver for the platform (${CLAUDE_PLUGIN_ROOT}/principles/08-testing-qa.md → "Drivers"): Playwright or the agent's browser tools for web; curl or the test client for an API.

Find the impact

Do the impact assessment in ${CLAUDE_PLUGIN_ROOT}/principles/08-testing-qa.md, "Tests for a change". Then:

  1. Run only the automated tests that the change can affect. Do not run all the tests, except when the assessment tells you to.
  2. Walk the acceptance criteria of the task (below).
  3. A changed shared file can affect other features. For each of these features, walk its main path one time, and its state that uses the shared file.
  4. Put the impact assessment in your report.

Walk it

  1. Acceptance criteria, one by one, as a customer would. Record what you did and what you saw.
  2. States: loading (throttle or pause the backend), error (stop the API so reads fail), offline (the browser offline plus the backend unreachable), empty (a fresh account).
  3. Unhappy paths: close the tab or kill the process mid-write or mid-upload, then reopen; a plan limit; permission denied; two users at once; a slow network.
  4. Accessibility: the largest text size (200% zoom, 320 px reflow), keyboard only, screen reader labels on new controls, reduce motion on. Reset each setting afterwards.
  5. Data: check the rows, files and jobs the action should create, with direct queries, counts and outcomes.
  6. Lifecycle: a reload, a second tab, the back button, a session that expires, a deploy while the page is open.

Scenarios

  • offline: warm the cache, go offline (network off and backend unreachable), reload: everything the product promises offline shows from cache. Make one of each write, close the tab, reconnect: each write lands exactly once. Measure the latency to the other user with a realtime listener.
  • layout: 320 px, a tablet, a desktop, 200% zoom, portrait and landscape. Screenshot each key screen.

Evidence and cleanup

  • Capture screenshots with capture-screenshots (states × sizes × settings, each one looked at), plus numbers (row counts, latencies).
  • Report each criterion as pass / fail / not verified (why). Never round "not verified" up to "pass".
  • File tickets for anything found that's outside the task's scope.
  • When another agent or session continues from here, write a handoff with these results.
  • Stop everything you started (dev server, browsers, local stack, listeners) and reset changed settings. Leave alone what you didn't start.

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.