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

demo

Skill2 starsChanged 7 days agoApache-2.0View on GitHub

Present a bug reproduction or finished test run to an audience — headed slow-motion replay, handover mode at page.pause(), screen recording only on explicit request. Use for "/demo <KEY-123>", "präsentier den Bug", "zeig die Repro", "lass mich probieren", "nimm das auf". Not for testing something new.

29 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use spec, implement and 3 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • SentrySkill

    spec

    Produce a verified spec — problem, desired outcome, and acceptance criteria — before building. Use when given a GitHub/Linear issue link, when starting from a fuzzy in-conversation idea, or asked to "figure out what to do" / "scope this" / "what should I build" — whenever the *what* isn't yet pinned down. Grills the ambiguities, checks the premise against the code, and produces acceptance criteria you sign off on — then hands to design-first.

    Also in 4 projects that use this

  • Hugging FaceSkill

    implement

    Make tests pass. Invoke after /write-tests produces failing tests.

    Also in 4 projects that use this

  • SentrySkill

    review

    Fixture review skill.

    Also in 4 projects that use this

  • AWSSkill

    ship

    Finalize dev work on an issue branch — update docs, push, and open a PR.

    Also in 3 projects that use this

  • NVIDIASkill

    research

    Before designing a feature, research how comparable libraries solve the problem, check best practices, known pitfalls, and upstream changes. Use when starting a new feature, investigating an unfamiliar area, or the user asks to research something.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 27 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

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

    highlight before action: every click/fill goes through a small helper that outlines the target element (~3 px solid outline) for ~600 ms, …

    line 73
  2. Distinctive:

    Keep it to the steps the audience will actually see — talking points, not a test protocol.

    line 125

“Common” is measured: kept by at least 6 in 10 of the 27 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

/demo — presenter mode

Show the blueprint to people: replay a bug reproduction or a finished test run visibly (headed browser, slow motion), hand the prepared browser over so the human can click through it themselves, and — only when explicitly asked — record the replay as .webm.

The presenter never tests. Everything shown comes from an existing, verified source: the bug ticket's reproduction steps or a finished run's evidence. If neither exists, say so and route there first (bug-reproduction for a bug without a confirmed repro, test-ticket for an untested feature) — a demo built on guesses is a demo of the wrong thing.

Modes

ModeTriggerWhat happens
1 · Present (default)/demo <KEY-123>, "zeig die Repro", "zeig wie du getestet hast"Headed slow-mo replay of the repro/test steps + demo card in chat
2 · Hands-on"lass mich probieren", "ich will selbst klicken"Replay prepares the state, stops at page.pause() — the human clicks on
3 · Recordonly explicit: --record, "nimm das auf"Same replay with video: 'on' → .webm into the demo dir

The most common run is a chain: present first (mode 1), then hand the same prepared browser over (mode 2) — "zeig es mir in der UI, dann lass mich probieren". When a request mentions both showing and trying, run the chain by default and say so up front.

Hard rules

  • Never start without the readiness click. The visible replay begins only after the user answered the clickable "Bereit?" question with "Starten" (below). A demo nobody noticed starting is a demo that failed.
  • Never record without an explicit request. No --record and no "nimm das auf" in the request = no video, no screenshots beyond what the replay itself needs, no silent capture of any kind. Asking "soll ich aufnehmen?" is fine; recording by default or "just in case" is a violation.
  • Secrets stay masked. Logins run through the Playwright --secrets mechanism exactly as in a test run — the audience never sees a value, and the demo card references users by role name (qa-standard), never by credential.
  • Stage only (existing rule — a demo is a test run in front of people).
  • No JIRA writes. The demo composes nothing, posts nothing. If the audience wants the recording attached to the ticket, that is a deliberate follow-up via report-to-jira — never part of the demo.
  • Delegation visibility applies double: the audience watches a screen, not the chat. Narrate what is about to happen before it happens (the demo card is that narration).

Steps

  1. Find the source. Given <KEY-123>: route to the project as usual (projects/<KEY>/PROJECT.md + product if routed), then locate, in order:
    • the ticket's own repro steps (bug tickets — read via atlassian MCP), or
    • the newest finished run in evidence/<KEY>/<TICKET>/ — execution_summary.json plus its screenshots (several runs → newest; if the user named one, use that). Extract the step list: numbered, one action per step, with the expected observation per step. Preflight applies: the pre-tool guard expects a fresh preflight.sh <KEY> pass before any browser call — a demo on an unreachable stage is the same false-positive factory as a test.
  2. Build the demo spec into the run's demo dir (evidence/<KEY>/<TICKET>/<ts>/demo/ — a fresh <ts> for the demo, not the original run's): a temporary Playwright CLI spec that replays the steps headed and slowly (slowMo, e.g. 500–1000 ms per action), with a code comment per step naming what the audience should notice. The scaffold always carries the visibility helpers — the audience must see where things happen, not just that something happened:
    • highlight before action: every click/fill goes through a small helper that outlines the target element (~3 px solid outline) for ~600 ms, then acts — a wrapper around locator.evaluate, no dependencies;
    • click dot: one page.addInitScript that draws a pulsing dot at every real pointer event, fading after ~1 s;
    • start banner: the spec's first step overlays a banner ("Demo startet — Schritt 1 folgt") for ~1.5 s, so a full-screen audience sees the beginning even without watching the chat. Keep it in the project's own test suite layout when one exists (its config already knows base URL and storage state); otherwise scaffold a minimal standalone config in the demo dir. Never write demo specs into the product's repo.
  3. Present (mode 1). Print the demo card first (see below), then ask readiness with the harness's clickable question UI (Claude Code: AskUserQuestion; opencode: the question tool) — options "Starten" / "Noch nicht", never a question that needs typing (plain-text fallback only when the harness has no question UI). Only on "Starten": run the spec headed, bring the window to the front (page.bringToFront()), let it settle ~1.5 s, then the start banner and step 1. Narrate step boundaries as they happen ("Schritt 3: der Coupon wird eingelöst — beobachte die Gesamtsumme"). When it ends: one sentence on what was shown and where the evidence lives.
  4. Hands-on (mode 2) — and the 1→2 chain. Same spec, but it stops at page.pause() right after the prepared state (logged in, right page, right test data). Print the demo card with the 3–5 clicks the human should try, say explicitly "das Fenster gehört dir — klick dich durch, ich warte", and wait. When the request chains modes ("zeig es mir, dann lass mich probieren"): present to the end first, then ask via the clickable question UI ("Übergeben — du klickst" / "Genug gesehen") and only on "Übergeben" run the handover. When the human is done, resume/finish the spec (cleanup only — no assertions about what the human did; their exploration is not a verdict).
  5. Record (mode 3 — only on explicit request). Re-run the same spec with video: 'on' in the config and save the resulting .webm into the demo dir as demo-<KEY-123>.webm. No conversion, no editing, no upload — the file is the deliverable. Say where it is.

The demo card

Print before every mode, in the user's language:

Demo: <KEY-123> — <one-line subject>
Quelle: <bug ticket repro | evidence run <ts>>
Modus:  <Vorführung | Selbst ausprobieren | Aufzeichnung>

1. <Schritt> → worauf achten: <Beobachtung>
2. …

Keep it to the steps the audience will actually see — talking points, not a test protocol.

Output to the user

Mode + demo-card path (it lives in the demo dir as demo-card.md), what was shown, and for mode 3 the .webm path. One short summary — the demo itself was the content.

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.