truongfelix/workflow/checks
Claude Code and agent skill by truongfelix/workflow
- community
Checking that the SKILL.md file is still in the repository.
Show and change how the code is checked before a merge in this project - by its own check (a command whose exit code says the code is ready, such as its tests, run on this machine), by GitHub CI, both, or nothing yet - with each repo's check command and the rules kept (a release needs a pass on the code it tags, each bug fix comes with a check that fails without it, the main branch is checked again after a merge). Use it when the user types /checks, asks how the code is checked or wants it chan…
Install
npx vskill@latest install truongfelix/workflow/checksSKILL.md
Checks
How the code is checked before a merge is the owner's choice, in .claude/ground-truth.conf: by the project's own check, its gate (GATES: each repo's command, run by gate.sh run, which keeps a receipt of each commit that passed), by GitHub CI (GITHUB_CI: the repos whose CI on GitHub counts), both, or nothing yet (both empty). The rules kept (GATE_RULES) hold for whichever checks a repo, and the board shows only what applies. A repo can keep other rules than the project's (REPO_RULES: its own list, which replaces GATE_RULES for that repo alone). This skill shows those lines and changes them with the owner's answers.
Ask as CLAUDE.md's rule "A question to the owner" says. The gate is "the check that says the code is ready" (its tests), never a bare "gate".
Steps
- Read
REPOS,GATES,GATE_PASS,GITHUB_CIandGATE_RULESandREPO_RULESin.claude/ground-truth.conf, then look in
each code repo for what it already has: a .github/workflows/ folder (GitHub CI), a test command or a local CI script (its own check). Outside setup, say in a few lines what checks each repo now and the rules kept (each as its sentence in step 4), then ask: keep it as it is, or change it; not when the owner has already said what changes (a repo attached later and its check, or "no check yet": step 2 with that answer).
- How is the code checked: "How is your code checked before a merge?", four options, the one the repos already
have recommended:
- Its own check: a command run on this machine, such as its tests; a release needs it to have passed here.
- GitHub CI: the checks GitHub runs on each push and pull request, read with
gh(installed and logged in). - Both: both must pass.
- Nothing yet: no check exists (a new project). No rule holds; run this skill again once there is one.
One answer for every code repo of a workspace (never its docs repo), unless the owner says they differ. A repo attached later is asked about alone, from this step on: the others keep their checks and their rules.
- Its own check, for each repo: the command found in step 1, in a question of its own: "Is
<command>the check
that says the code is ready?" (another one is typed in the chat). A local CI script the project already has is that command: from then on it runs through gate.sh run, never beside it, so one run gives both records, and the workflow never reads the script's own. A script that exits 0 once it has run, green or not (stages skipped, a partial run), says it passed in a line it prints: ask "How do I know it passed: its exit code, or a line it prints?", and set that line in GATE_PASS.
- The rules (not with nothing yet), one question where the owner picks any or "None of these", each rule
(receipt, regression, merge) in the plain words rule_text gives in .claude/skills/session-start/lib.sh, as the board's Settings tab shows them. Name no preset: Production, Steady and Prototype are what the ground truth and the board call the result. This is the project's rules, for every repo. Then, when the project has more than one code repo: "Does any repo keep other rules than these?", and for each one the owner names, the same question for that repo alone; a repo the owner does not name follows the project's. A repo the owner changes later is changed alone. Two repos keep no rule, with no question, since nothing checks them: the docs repo of a workspace (the docs in a repo of their own), which holds no code, and a repo answered Nothing yet while other repos are checked, until its first check.
- Write the four lines, each repo by its name in
REPOS:GATES=("api|make check"), `GATE_PASS=("api|^All
checks passed"), GITHUB_CI=("api"), GATE_RULES="receipt regression merge"; nothing yet, for every repo, empties all four. A repo with rules of its own gets an entry in REPO_RULES=("docs|" "web|regression merge") (an empty list: it keeps none), and so do the two of step 4 that keep no rule: the docs repo of a workspace, and a repo with nothing yet, which has no line in GATES or GITHUB_CI. A repo given the project's rules again loses its entry, as does a repo at its first check. With receipt kept for a repo (its own rules, else the project's), run bash .claude/skills/session-start/gate.sh hook <repo>` for each such repo checked. Exit 1: the repo has a pre-push hook of its own; add the lines printed to it, with the owner's yes.
- Verify:
bash .claude/skills/session-start/ground-truth.sh --no-fetch, its section== checks. A ⚠ NO
RECEIPT goes away with gate.sh run <repo> on the default branch, once committed (a receipt is for a commit: at setup, after the setup's commit); a ⚠ CI that is not green is news for the owner; an UNKNOWN on CI means gh cannot read it (not installed, not logged in, no GitHub repo): fix that, or drop GitHub CI for that repo.
- Say what changed in plain words, and commit
.claude/ground-truth.confas the row "Commit the docs" says,
then push as "Push the docs" says. At setup, the setup commits it with the rest.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingtemplate/dot-claude/skills/checks/SKILL.md
Popularity
- Repository stars
- 65
- Repository forks
- 6
- 7-day trend score
- 0
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/truongfelix/workflow/checks)