← ClaudeAtlas

checklisted

Run a devteam project's mechanical checks — traceability, references, report blocks, scopes — read what each result means, commit through the gate that runs them on the commit itself, and accept a finding by a decision. Run on every report before a verifier runs, before closing any task, and whenever a result or a refusal needs reading.
alternative-intelligence-cp/claude-skills · ★ 1 · AI & Automation · score 74
Install: claude install-skill alternative-intelligence-cp/claude-skills
# The checks **Every hole this discipline has found was found by a check that diffs two lists, and none of them by a test.** These are those diffs. They are scripts with exit codes, not careful readings, and each ships a negative control beside it because *a check that has never failed has not been shown to work* (P-35). ```bash python3 "${CLAUDE_PLUGIN_ROOT}/scripts/check_trace.py" <project> python3 "${CLAUDE_PLUGIN_ROOT}/scripts/check_refs.py" <project> python3 "${CLAUDE_PLUGIN_ROOT}/scripts/check_report.py" <project> T-n[.S-m] python3 "${CLAUDE_PLUGIN_ROOT}/scripts/check_scope.py" <project> [T-n[.S-m]] python3 "${CLAUDE_PLUGIN_ROOT}/scripts/run_controls.py" # prove the checks ``` ## What a result means **Every other skill points here for what a check's exit code means.** The contract's home is `scripts/result.py`, and the plugin's `templates/FORMATS.md` §"What each check reads" states it. | Exit | Result | What to do | |---|---|---| | `0` | clean | Everything the check owns was evaluated, and nothing was found. **This is the only pass** | | `1` | findings | Fix what each line names. The line also names any part the check did not evaluate | | `2` | could not run | The fault is in the invocation — its arguments, no repository, no `devteam/` — and not in the project. Fix the command | | `3` | not evaluated | No finding, and at least one part the check did not look at, each named with its reason: a row its grammar could not read, a piece of an identifier fiel