# Verify
> Run moon typecheck, lint, and test tasks for affected projects after edits. Use only in a configured moon workspace; report fresh command evidence without adding review or handoff ceremony.
- Skill: `g-bastianelli/verify` (Agent Skill)
- Install (CLI): `npx skillmds@latest add g-bastianelli/verify`
- Raw SKILL.md: https://api.skillmd.com/api/skills/g-bastianelli/verify/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: g-bastianelli (https://skillmd.com/u/g-bastianelli)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/g-bastianelli/verify
---
# verify
## Voice
Read `../../persona.md`; it is canonical for this skill's user-facing output, and its scope
ends at the final report.
## Outcome
Verify the current change through moon's affected graph and report the real result. Execute
the checks directly: do not delegate, run an additional code-review pass, write a ledger, or
present commit/PR menus.
## Workflow
1. Walk upward to the nearest real workspace config. It is a
`workspace.{json,jsonc,hcl,pkl,toml,yml,yaml}` file under `.moon/` or `.config/moon/`.
A bare `.moon/` directory is not a workspace. If none exists, say the skill does not
apply and stop; repository-native verification belongs to the repository's normal
workflow.
2. Run `moon --version` and `git status --porcelain` from the workspace root. Build the
affected project set by piping changed-file JSON into Moon's project query:
```sh
moon query changed-files | moon query projects --affected --downstream deep
```
This compares the configured default branch with `HEAD`, so it also works on the clean
worktree required before a PR. When the worktree is dirty, additionally run the same
pipeline with `moon query changed-files --local` and union both project sets. Honor an
explicit revision request with `--base --head ` on the first query. If moon
is unavailable or the combined set is empty, report that no verification ran; do not
claim success.
3. From the project-query JSON, collect every project id, including downstream projects,
and the `typecheck`, `lint`, and `test` tasks each project actually defines. Unless the
user requested narrower checks, run the applicable targets explicitly for every project,
for example:
```sh
moon run app-a:typecheck app-a:lint app-a:test app-b:typecheck app-b:test
```
Never rely on `: --affected --downstream deep` to cover downstream projects: moon
may omit their tasks unless task dependencies connect them. Trust moon's cache. Do not
replace moon tasks with raw package-manager, compiler, linter, or test-runner commands.
4. Report the exact targets and a per-project pass/fail summary. Name affected projects with
no applicable task. Include the relevant failing output verbatim. Verification succeeds
only when every applicable target ran after the last edit and exited successfully;
otherwise state what remains unverified.
Verification is read-only apart from build and test outputs created by moon. Never edit
source files, commit, push, or rebase from this skill.