Skip to content
Back to skills

Verify

ASecurity

Verifies a change actually works by running it and quoting the observed output — a build passing is an assertion, a quoted test summary is evidence.

  • 4 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added September 26, 2026
ai-agentscode-review

Security analysis

A100/100

Scanned September 26, 2026

npx -y skills add xiaolongnk/SuperHarness --skill verify --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Verify?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Verify
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/xiaolongnk-verify/badge)](https://www.skillsdirectory.com/skills/xiaolongnk-verify)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: verify
description: Verifies a change actually works by running it and quoting the observed output — a build passing is an assertion, a quoted test summary is evidence.
domains: [quality, verification]
tier: both
---

# Verify skill

## When to use

Invoke before claiming any change is **fixed / passing / done**, and before committing or shipping.
This is the gate that stops "I think it works" from being reported as "done."

Always run it — there's no change too small to confirm. For a pure behavior change, also confirm the
behavior in the running app, not just a passing build.

## Steps

1. **State what "works" means** for this change — the specific observable outcome that proves it
   (a test passes, a command returns X, the feature behaves correctly in the running app). Pull this
   from the plan's definition of done if there is one.
2. **Run the project's checks** — lint, then the test suite (scoped to the area you changed is fine,
   but don't skip the relevant tests). Run the real command, not a remembered result.
3. **Read the output.** A command that *errored* is not a pass — confirm it exited cleanly and the
   numbers are real (e.g. `Tests 3 passed`, not a crash or a "0 tests run"). A green-looking line
   from a run that never executed your test is a false pass.
4. **For a behavior change, exercise the behavior.** Start the app, drive the actual path the change
   affects, and observe the result. Reproduce the original failure scenario if this was a bug fix —
   confirm it no longer happens.
5. **Verify the artifact that actually runs**, not just the working tree. A fix that's saved but not
   in the running build/deploy is not shipped — confirm the change is present in what executes.
6. **Quote the evidence in your report.** Paste the test summary line, the command output, or
   describe the observed behavior. Claiming success without quoting what you saw is the failure
   this skill exists to prevent.
7. **If it doesn't work, say so plainly** and stop — do not report done. Hand the failing output to
   `/code-review` or back to the fix.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.