Implements a deterministic testing workflow by defining acceptance criteria, running failing checks, and iterating until tests pass.
- GitHub stars
- 5
- Files
- 4
- Last commit
- Oct 29, 2026
- License
- MIT
- Allowed tools
- —
What does the verify-loop skill do?
verify-loop implements a deterministic testing workflow by defining acceptance criteria, running failing checks, and iterating until tests pass.
Summary generated from the skill's SKILL.md.
About the verify-loop skill
Implementing a deterministic testing workflow by defining acceptance criteria and iterating through failing checks is what this skill enables. Developers use it when starting a new implementation or refining a fix to ensure that every change is validated by a concrete, repeatable test. Instead of relying on memory or manual error copying, an agent uses this skill to run specific commands and observe the direct output of a runner.
An agent should load this skill to manage the cycle of running a check, confirming it is red, and repeating the process until it reaches a green state. It relies on external runners like pytest, compilers, or performance traces provided by the repository. The skill includes specific wrappers such as verify.py for Python, verify.sh for POSIX environments, and check-snapshot-console.cjs for Playwright-based snapshot and console error checks. It requires that the user provides the actual command and fresh runner output to maintain a single source of truth for the task.
What an agent can do with it
- 1Verifying web interface changesUse a snapshot and console error check to ensure a page renders correctly without installing new packages.
Written from the skill's SKILL.md. Every use case is checked against a quote from it.
How to install the verify-loop skill
Each command copies the skill folder into the client's skills directory. Restart the client or start a new session to load it.Install the verify-loop skill in Plugin
/plugin marketplace add KunanonJ/superstacks /plugin install superstacks@superstacks
Install the verify-loop skill in Claude Code
~/.claude/skills/verify-loop/SKILL.md
$ npx degit KunanonJ/superstacks/plugins/superstacks/skills/verify-loop ~/.claude/skills/verify-loopInstall the verify-loop skill in Claude Code (project)
.claude/skills/verify-loop/SKILL.md
$ npx degit KunanonJ/superstacks/plugins/superstacks/skills/verify-loop .claude/skills/verify-loopInstall the verify-loop skill in Codex
~/.agents/skills/verify-loop/SKILL.md
$ npx degit KunanonJ/superstacks/plugins/superstacks/skills/verify-loop ~/.agents/skills/verify-loopInstall the verify-loop skill in Gemini CLI
~/.gemini/skills/verify-loop/SKILL.md
$ npx degit KunanonJ/superstacks/plugins/superstacks/skills/verify-loop ~/.gemini/skills/verify-loopInstall the verify-loop skill in Claude.ai
npx degit KunanonJ/superstacks/plugins/superstacks/skills/verify-loop verify-loop && zip -r verify-loop.zip verify-loop # claude.ai → Customize → Skills → + → Create skill → Upload a skill → verify-loop.zip
Automated checks
Ran Nov 16, 2026Full report →Passed our automated checks. That is not a security audit: read the scripts before you let an agent run them. How checks work
- Source repository linked and reachablekunanonj/superstacks/plugins/superstacks/skills/verify-loop is public and reachablegithub.com/kunanonj/superstacks/tree/HEAD/plugins/superstack ↗Pass
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit 2026-10-29github.com/kunanonj/superstacks/commits ↗Pass
- No download-and-run or credential theft in scripts3 scripts read; no download-and-run, remote eval, base64 execution or credential uploadPass
- No hidden instructions in SKILL.mdNo override phrases, safeguard switches, credential requests or hidden charactersgithub.com/kunanonj/superstacks/blob/HEAD/plugins/superstack ↗Pass
- Pre-approved tools are scopedNo pre-approved tools; the agent asks before using onegithub.com/kunanonj/superstacks/blob/HEAD/plugins/superstack ↗Pass
- SKILL.md follows the Agent Skills formatValid name and description; says when to use it; 44 linesgithub.com/kunanonj/superstacks/blob/HEAD/plugins/superstack ↗Pass
SKILL.md
View on GitHubVerify loop
The engineer designs a runnable check. The model spends tokens on judgment, not on being a human who copies errors between tools.
Before writing code
- Name the acceptance criteria in one sentence.
- Pick a deterministic check that would fail today and pass when the work is done: unit or integration test, CLI assertion, perf trace, or a snapshot/console script.
- Copy a runner from
scripts/next to this skill into the repo (or run it in place). Do not install packages. - Run the check. Confirm it is red. Paste the runner output (
VERIFY_FAIL). - Change code. Re-run the same command. Repeat until
VERIFY_PASS. - Hand the same command to
verification-before-completionbefore any success claim.
Never retype compiler, test, browser, or linter errors from memory. Re-run the command.
Runners
| File | When |
|---|---|
scripts/verify.py |
Python stdlib wrapper. Prints pass/fail evidence. |
scripts/verify.sh |
POSIX wrapper with the same evidence format. |
scripts/check-snapshot-console.cjs |
Page snapshot plus console-error check. Uses Playwright already in the project. Exits 2 if it is missing. Never installs it. |
python <this-skill>/scripts/verify.py -- <project-check>
bash <this-skill>/scripts/verify.sh -- <project-check>
node <this-skill>/scripts/check-snapshot-console.cjs <url> [shot.png]
Replace <project-check> with the repo's real command (pytest, a compiler, a perf trace). The wrapper does not choose the check.
Rules
- One command is the source of truth for this ticket. Do not swap it mid-loop without saying why.
- Red before green. A check that was never red does not prove the fix.
- If the check needs a browser, use the snapshot template or a CDP script the repo already runs. Do not add runtime installs.
- Paste fresh runner output. Screenshots without
VERIFY_PASS/VERIFY_FAILare not enough.
Files in this skill
4verify-loop skill questions
How do I install the verify-loop skill in Claude Code?
Run npx degit KunanonJ/superstacks/plugins/superstacks/skills/verify-loop ~/.claude/skills/verify-loop. Claude Code loads skills from ~/.claude/skills (all projects) or .claude/skills (one project) and uses a skill when its description matches the task.
Which clients can use the verify-loop skill?
Any client that reads the Agent Skills format: Claude Code, Codex, Gemini CLI and claude.ai (upload as a ZIP). Scripts in the skill run only where the client allows code execution.
What is the required sequence for the verify loop?
The agent must define acceptance criteria, pick a deterministic check, run it to confirm it fails (red), iterate on the code, and repeat until it passes (green).
How can I verify web pages using snapshots?
Use the check-snapshot-console.cjs script which utilizes Playwright already present in the project to check page snapshots and console errors.