# Agent Md Verify

> Use when finishing work in a repository that has agent-md installed; runs the declared checks, updates memory/progress.md, and reports concrete verification evidence instead of self-grading.

- Skill: `iamfakeguru/agent-md-verify` (Agent Skill)
- Install (CLI): `npx skillmds@latest add iamfakeguru/agent-md-verify`
- Raw SKILL.md: https://api.skillmd.com/api/skills/iamfakeguru/agent-md-verify/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: iamfakeguru (https://skillmd.com/u/iamfakeguru)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/iamfakeguru/agent-md-verify

---


# agent-md Verification

Use this skill before claiming work is complete in an agent-md repository.

1. Read `memory/verify.md` and identify the checks required for this task.
2. Run the project verification commands declared in `agent-md.toml` when present.
3. If `agent-md.toml` is absent, use the repository's existing test/lint/typecheck scripts or the hook heuristics.
4. For source changes, update `memory/progress.md` with the completed atomic task and the checks that passed.
5. If no checks exist, say the work is unverified and add a follow-up item to define checks.
6. Do not describe code as correct based only on inspection. Provide command output, artifact paths, or reviewer evidence.

Useful helper:

```bash
./.agent-md/bin/discover_helpers.sh
```