Skip to content
Back to skills

Verify

ASecurity

Verify a change actually does what it claims by running it and observing behavior

  • 10 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 19, 2026
ai-agentsshellgit

Works with

  • cli

Security analysis

A100/100

Scanned September 19, 2026

npx -y skills add initxy/noeta --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/initxy-verify/badge)](https://www.skillsdirectory.com/skills/initxy-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: Verify a change actually does what it claims by running it and observing behavior
argument-hint: [what to verify, e.g. "the new login retry logic"]
---

# Verify a change

Confirm that a code change truly does what it's supposed to — by running it and
observing real behavior, not just by reading the diff.

Take the claim or feature under test from the conversation.

## Steps

1. Pin down the claim. Run `git_diff` and `read` on the touched code to state, in
   one sentence, what behavior this change is supposed to produce.
2. Find how this project runs and tests. `read` the manifest / `Makefile` / CI
   config and `grep` for existing test commands so you use the project's real entry
   points rather than inventing them.
3. Exercise it with `shell_run`:
   - Run the relevant automated tests (whole suite or a targeted single test) and read
     the output.
   - If behavior is observable another way (a CLI command, a script, a smoke run),
     execute that too and capture the actual output.
4. Compare observed behavior against the claim from step 1. Cover the happy path and at
   least one edge / failure case the change should handle.
5. Report a clear verdict: **verified** or **not verified**. Quote the exact command(s)
   you ran and the output you observed. If it fails, point to the smallest reproducing
   command and the precise mismatch between expected and actual.

Do not "fix" the code here — only verify and report. If you had to change anything to
make it runnable, call that out explicitly.

(fork agent: general-purpose)

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…