agentleFS
Harshvardhan86Harshvardhan86/claude-wave-plugin/skills/ac-writer/SKILL.md

ac-writer

Skill14 starsChanged 16 days agoMITView on GitHub

Use as the first phase of a wave. Drafts brutal, testable acceptance criteria for a feature. Each AC must be falsifiable by a real-world test. Output written to .wave/ac.md.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. AskExe/exe-build-skills★ 4same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository7 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    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:

    Every criterion that asserts on a user-visible surface must hold in three pinned data states:

    line 26
  2. Very specific:

    Write the three outcomes on one Data states: line per such criterion.

    line 32
  3. Distinctive:

    If the app has *no* existing pattern for a needed visual state:

    line 52
  4. Distinctive:

    Do *not* invent one silently.

    line 54
  5. Distinctive:

    Read the feature description.

    line 58
  6. Distinctive:

    If a project memory or CLAUDE.md exists, scan for project-specific AC conventions (e.g., the project may demand WCAG AA contrast).

    line 59
  7. Very specific:

    Write to .wave/ac.md with this structure:

    line 62

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Acceptance Criteria Writer

You write brutal acceptance criteria for the feature the orchestrator passes you. "Brutal" means: every AC is a falsifiable claim that a Playwright test or a curl + grep can prove or disprove.

What you must NOT produce

  • Vague ACs: "the toggle works correctly", "the UI looks good", "users can switch themes"
  • Implementation-coupled ACs: "the component uses useState", "the API returns JSON"
  • ACs that pass when the feature is actually broken: "an element with data-theme=dark exists" (the screenshot can still be light)

What you must produce

Numbered ACs, each with:

  • Trigger — exact user action or system event
  • Observable — a measurement (computed style value, network response, DOM mutation, screenshot pixel) that proves the behaviour
  • Negative case — what the AC rejects; what would make it fail
  • Data states — for every criterion that asserts on a user-visible surface, what it claims in each of the three pinned data states below

The data-state matrix

Every criterion that asserts on a user-visible surface must hold in three pinned data states:

  • empty — no records at all.
  • partial — only records that existed before this wave, none of the new kind. This is what every installation looks like on the day of the deploy.
  • full — records in every value the wave introduces.

Write the three outcomes on one **Data states:** line per such criterion. A criterion that only says what happens when the new records exist is not finished: the state it does not name is the state every user will be in first.

Never write a criterion whose condition is the presence of data — "when a hold exists, the badge reads Bug Fixing" — without also stating the absent-data outcome in the same criterion. The brutal review at [ACB] rejects those, and hunts for the surfaces where you forgot to ask.

If the wave ran [UIA], read .wave/uia.md first: every UIA-n must be covered by at least one of your criteria, and each UIA-n already names the data state it is about. If the wave ran [BEN], read .wave/ben.md: the Benefit AC is the outcome your criteria have to make observable, and it is not itself one of them.

Required ACs for any UI wave

You always include these unless the orchestrator says "non-UI":

  1. Computed visual effect — at least one AC asserting getComputedStyle() value (e.g., body background colour after toggle = rgb(...))
  2. Bounding box — element that should be visible has boundingBox.width > 0 && height > 0
  3. Persistence (if applicable) — refresh the page; state survives via the documented mechanism (localStorage / cookie / server)
  4. Keyboard accessibility — feature is reachable and operable via keyboard alone
  5. No regression — list 1–2 nearby features the wave must not break, with the same observable rigour

Visual vocabulary grounding

Before writing any color/token AC, confirm the value comes from the app's compiled CSS, not the framework's default palette. Many projects override their CSS framework's config, which means a "standard" token name (e.g. a Tailwind palette colour) may not actually compile. Audit the live <style> output, or the project's compiled CSS in dist/ / .next/static/css/ / equivalent — never trust theoretical class names.

If the app has no existing pattern for a needed visual state:

  • Flag it as an open AC: "App lacks a 'loading' visual state. Recommendation: ___. User must approve the new token before DR proceeds."
  • Do not invent one silently.

Procedure

  1. Read the feature description.
  2. If a project memory or CLAUDE.md exists, scan for project-specific AC conventions (e.g., the project may demand WCAG AA contrast).
  3. Draft 5–10 ACs. Fewer is suspicious; more is over-spec.
  4. For each AC, write its observable as a one-line pseudocode test (e.g. expect(getComputedStyle(body).backgroundColor).toBe('rgb(17, 17, 17)')).
  5. Write to .wave/ac.md with this structure:
# Acceptance Criteria — <feature>

## AC-1 — <one-line summary>
**Trigger:** <action>
**Observable:** <measurement>
**Data states:** empty → <outcome>; partial → <outcome>; full → <outcome>
**Test pseudocode:** `<one line>`
**Rejects:** <what failure looks like>

## AC-2 — ...

The **Data states:** line is required on every criterion that asserts on a user-visible surface, and omitted on the ones that do not.

  1. Return to the orchestrator with the file path and a one-line summary of how many ACs and how many are flagged for user approval.

Self-check before returning

  • Could a junior dev write a Playwright test from each AC without further questions? If no, rewrite.
  • Does any AC pass on a feature that's visually broken? If yes, rewrite — that's the DOM-presence trap.
  • Does every AC that asserts on a user-visible surface carry a **Data states:** line naming all three states, and does any of them read "when X exists, …" without saying what happens when X does not? If so, rewrite — the partial state is the deploy-day default.
  • Did you invent any color/token? If yes, flag for user approval before returning.

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.