agentleFS
roronoazoroshao369roronoazoroshao369/vibe-coding-os/skills/core/verification-before-done/SKILL.md

verification-before-done

Skill1 starsChanged 5 months agoMITView on GitHub

Make completion claims only after relevant checks, review, and limitations are explicit. Use before saying done, opening a PR, merging, handing off, or ending a long work session. Also use before final response, handoff, merge, or deployment. Task goal, acceptance criteria, changed files, validation commands, test output, known limitations, and environment constraints. When verifying a change that runs in a browser, app, or service, check the 5 axes of runtime behavior. This pattern is framework-agnostic; apply the axes that match…

3 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

5 rules. 0 are common, 4 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 4

    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. Distinctive:

    Use before saying done, opening a PR, merging, handing off, or ending a long work session.

    line 20
  2. Distinctive:

    When verifying a change that runs in a browser, app, or service, check the 5 axes of runtime behavior.

    line 29
  3. Very specific:

    Claude Code (browser): use chrome-devtools MCP for DOM/console/network/visual.

    line 40
  4. Distinctive:

    Run the smallest relevant checks first, then broader validation when feasible.

    line 46
  5. Distinctive:

    Do not convert environment limitations into success claims.

    line 50

“Common” is measured: kept by at least 6 in 10 of the 3 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

Skill: Verification Before Done

Purpose

Make completion claims only after relevant checks, review, and limitations are explicit.

When to use

Use before saying done, opening a PR, merging, handing off, or ending a long work session. Also use before final response, handoff, merge, or deployment.

Inputs

Task goal, acceptance criteria, changed files, validation commands, test output, known limitations, and environment constraints.

5-axis runtime verification (abstract)

When verifying a change that runs in a browser, app, or service, check the 5 axes of runtime behavior. This pattern is framework-agnostic; apply the axes that match your runtime.

AxisWhat to verifyEvidence
DOM / stateThe rendered output matches the spec (or pre-change baseline).Screenshot, DOM diff, snapshot test.
Console / logsNo new errors, warnings, or unhandled rejections.Console capture, log diff.
Network / APIThe expected requests fire with the expected payloads; responses match the spec.Network trace, API contract test.
PerformanceLCP / INP / CLS / TTFB / API p99 are within "Good" thresholds.CWV report, perf budget gate.
VisualThe UI renders correctly across the supported viewports / themes.Screenshot diff across breakpoints.

For multi-CLI projects, the 5 axes map to specific tools per CLI:

  • Claude Code (browser): use chrome-devtools MCP for DOM/console/network/visual; Lighthouse for performance.
  • Other CLIs (non-browser): substitute network/API/console equivalents; visual axis becomes "output file diff" or "headless screenshot".

Workflow

  1. Map acceptance criteria to concrete evidence: tests, validation, inspection, or user confirmation.
  2. Run the smallest relevant checks first, then broader validation when feasible.
  3. Record exact commands and outcomes, including failures and warnings.
  4. Apply the 5-axis runtime verification for any user-facing or latency-sensitive change.
  5. Inspect the final diff for scope, secrets, attribution, generated files, and stale notes.
  6. Do not convert environment limitations into success claims.
  7. End with a clear status: passed, failed, blocked, or partially verified.

Outputs

A verification report with exact commands, outcomes, evidence coverage, 5-axis matrix, limitations, and final readiness status.

Failure modes

  • Reporting tests as passed when they were not run.
  • Claiming success without running checks.
  • Using broad validation to hide a missing targeted check.
  • Ignoring failed checks because the patch looks correct.
  • Skipping one of the 5 axes on a user-facing change.
  • Omitting environment limitations.
  • "It works on my machine" — local verification is not p75 user verification.

Verification checklist

  • Acceptance criteria have matching evidence.
  • Exact commands and outcomes are listed.
  • Failures or limitations are visible.
  • Final status is not overstated.
  • All npm run validate:* commands invoked in this change exited 0 (record exit codes in commit body).
  • DOM/state axis captured: screenshot, DOM diff, or snapshot test attached for any UI change.
  • Console/logs axis captured: console.log/error/warn output recorded, zero new unhandled rejections.
  • Network/API axis captured: request/response trace or contract test recorded for any API change.
  • Performance axis captured: LCP / INP / CLS / TTFB / API p99 numbers recorded for any latency-sensitive change.
  • Visual axis captured: screenshot diff across the supported breakpoints (mobile, tablet, desktop).
  • Each check is recorded as PASS / FAIL / LIMITATION: — no "looks good" or "feels right".

Related skills

This skill is the evidence bar that backs the goal-driven member of the four-part engineering discipline set (think-before-coding, simplicity-first, surgical-changes, goal-driven):

  • skills/core/goal-driven-execution/SKILL.md — turns an imperative into a verifiable goal whose success condition is checked here.
  • skills/prompts/karpathy-engineering-discipline/SKILL.md — Think Before Coding and Surgical Changes.
  • skills/prompts/anti-overengineering/SKILL.md — Simplicity First.

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.