verify
Verification ladder. Pick smallest scope, run scripts in order, smoke UI, report PASS/NO-GO.
Best use case
verify is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Verification ladder. Pick smallest scope, run scripts in order, smoke UI, report PASS/NO-GO.
Teams using verify should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/verify/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How verify Compares
| Feature / Agent | verify | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Verification ladder. Pick smallest scope, run scripts in order, smoke UI, report PASS/NO-GO.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# Verify ## Purpose Provide a single, consistent way for agents to verify changes before claiming “done”. ## Canonical rule Verification is skill-only. - Do not rely on a global `verify` command. - Do not require other docs to define verification logic. This skill is self-contained. ## When to run Run verification for every change that alters behaviour, build outputs, tests, types, linting, packaging, CI, or UI. If you cannot verify, return **NO-GO** with a concrete reason and the smallest unblock request. ## Procedure ### 0) Pick scope (don’t waste time) - Prefer the smallest scope that covers what changed: - single package/app if changes are local - repo-wide only if shared libraries, configs, or multiple packages changed - If unsure, start scoped, then widen if failures suggest shared impact. ### 1) Discover the available verification scripts (don’t guess) 1) Identify the relevant `package.json`: - repo root for repo-wide checks - the changed package/app directory for scoped checks 2) List scripts from the relevant `package.json`: - Look for canonical script names first: - `verify` - `test`, `test:unit`, `test:ci` - `lint` - `typecheck`, `tsc` - `build` - `format`, `fmt`, `prettier:check` - `check` (sometimes bundles lint/typecheck/test) 3) If none exist, search one level up/down for where scripts live (monorepo layouts vary). If you still cannot find scripts, return **NO-GO** and ask for expected verification steps or for a `verify` script to be added. ### 2) Run verification in this order (stop on red) Run the best available equivalents in this order: 1) Format / basic hygiene (only if a check script exists) - Prefer `format:check` / `prettier:check` / `fmt:check` - Do not run auto-format as “verification” unless explicitly requested 2) Lint - Run the repo/package lint script 3) Typecheck - Run `typecheck` / `tsc` / equivalent 4) Tests - Run the smallest relevant test suite: - `test` / `test:unit` for code changes - include integration/e2e only when the repo provides it and it’s relevant 5) Build - Run `build` when: - you changed build outputs, bundling, env wiring, dependencies, or UI - or when the repo commonly relies on build as a gate Notes: - If the repo provides a single `verify` or `check` script, prefer it first. If it fails, fix before running anything else. - In monorepos, use the workspace/package targeting mechanism that exists in the repo (don’t assume pnpm). If unclear, run within the package directory. ### 3) UI smoke testing (required for UI/user-flow changes) If changes affect UI, routing, auth flows, forms, state, client behaviour, styling that could break layout, or anything user-visible, do a smoke test. #### 3.1 Start the app - Use the repo’s normal dev/start script for the relevant app (e.g. `dev`, `start`, `preview`). - If there are multiple apps, start the one you changed. If you cannot start the app due to missing env/secrets, return **NO-GO** and request the minimal env needed (or a mock mode). #### 3.2 Smoke checklist (happy path + one sad path) Do the minimum that proves the change works: - Load the page/route you touched - Exercise the changed behaviour end-to-end - Confirm no obvious runtime errors (console/network) and the UI behaves as intended - Do one “sad path” where relevant (invalid form submit, unauth access, empty state) #### 3.3 Basic a11y spot-check - Keyboard: tab order makes sense, focus is visible, primary actions reachable - Forms: error message appears, first invalid field gets focus (or is clearly indicated) - No colour-only signalling for errors or status - Reduced motion: no jarring animations if the change touches motion (best-effort) #### 3.4 Evidence Prefer using the `agent-browser` skill for smoke testing when real interaction matters. Capture evidence where possible: - screenshot of the critical state, or - short description of the steps performed + observed outcome ### 4) If something fails - Keep the fix small and local first. - If the failure is due to missing scripts or repeated workflow pain, recommend adding a `verify` script that bundles lint/typecheck/test/build for that scope. ## Output contract (mandatory) Include a section named `Verification` with either: ### PASS - `PASS:` - Commands run (exact) - Results (pass) - UI smoke (if applicable): steps + outcome + evidence note ### NO-GO - `NO-GO:` - What blocked verification (exact) - What you need to proceed (smallest unblock request) - What you did manage to verify (if anything)
Related Skills
skill-creator
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
modular-skills-architect
Map and refactor an agent context ecosystem: skills, commands/workflows, hooks, agent files, AGENTS.md templates, and docs. Output system map, module/dependency design, Register updates, and a concrete split/consolidate/rename/delete plan. Use when routing or ownership is messy.
heal-skill
This skill should be used when fixing incorrect SKILL.md files with outdated instructions or APIs.
create-agent-skills
Expert guidance for creating, writing, and refining Claude Code Skills. Use when working with SKILL.md files, authoring new skills, improving existing skills, or understanding skill structure and best practices.
agent-native-audit
Comprehensive agent-native architecture audit with scored principles and multi-slice review. Use for system-wide health checks or periodic audits.
write-judge-prompt
Design LLM-as-Judge evaluators for subjective criteria that code-based checks cannot handle. Use when a failure mode requires interpretation (tone, faithfulness, relevance, completeness). Do NOT use when the failure mode can be checked with code (regex, schema validation, execution tests). Do NOT use when you need to validate or calibrate the judge — use validate-evaluator instead.
validate-evaluator
Calibrate an LLM judge against human labels using data splits, TPR/TNR, and bias correction. Use after writing a judge prompt (write-judge-prompt) when you need to verify alignment before trusting its outputs. Do NOT use for code-based evaluators (those are deterministic; test with standard unit tests).
generate-synthetic-data
Create diverse synthetic test inputs for LLM pipeline evaluation using dimension-based tuple generation. Use when bootstrapping an eval dataset, when real user data is sparse, or when stress-testing specific failure hypotheses. Do NOT use when you already have 100+ representative real traces (use stratified sampling instead), or when the task is collecting production logs.
evaluate-rag
Guides evaluation of RAG pipeline retrieval and generation quality. Use when evaluating a retrieval-augmented generation system, measuring retrieval quality, assessing generation faithfulness or relevance, generating synthetic QA pairs for retrieval testing, or optimizing chunking strategies.
eval-audit
Audit an LLM eval pipeline and surface problems: missing error analysis, unvalidated judges, vanity metrics, etc. Use when inheriting an eval system, when unsure whether evals are trustworthy, or as a starting point when no eval infrastructure exists. Do NOT use when the goal is to build a new evaluator from scratch (use error-analysis, write-judge-prompt, or validate-evaluator instead).
error-analysis
Help the user systematically identify and categorize failure modes in an LLM pipeline by reading traces. Use when starting a new eval project, after significant pipeline changes (new features, model switches, prompt rewrites), when production metrics drop, or after incidents.
build-review-interface
Build a custom browser-based annotation interface tailored to your data for reviewing LLM traces and collecting structured feedback. Use when you need to build an annotation tool, review traces, or collect human labels.