← ClaudeAtlas

confab-assertion-auditlisted

Use when the user asks whether their test suite would actually catch bugs, wants a mutation-testing pass over target source and test files, asks to check for weak or tautological assertions, or wants confab's assertion audit run. Not for proving that a specific change, fix, wire, or numeric claim is correct -- use andon:andon-verify for adversarial verification of a change; this skill judges only the tests. Every finding is labeled real-tool or llm-reasoned.
Anselmoo/werkstoff · ★ 1 · Testing & QA · score 67
Install: claude install-skill Anselmoo/werkstoff
Check whether the given test files would actually catch plausible mutations to the given target source files. This skill's Verify phase is NOT optional — there is no `--skip-verification` flag on this domain's writer script, and `scripts/assertion_audit.py` requires a `--verify-json` argument it will refuse to run without. Do not look for a way to make this run faster by skipping it; the capability does not exist. ## Steps 1. Determine `repo_root`, the target source files, the test files, and (optionally) a named real mutation tool (e.g. `mutmut`) the user wants used. If the user didn't name one, still ask the agent to check whether one is available on PATH before falling back. 2. Optionally build (or reuse, if `confab-cycle` already built one this invocation — see "Shared symbol index" below) a symbol index of the target files for the agent to use as evidence. 3. Dispatch the `assertion-auditor` agent in **Find mode**: give it the target files, test files, the named tool (if any), and the symbol index (if built) — if step 2 was skipped, tell the agent no symbol index exists and it should read the target files directly instead. Ask for `{"findings": [...]}` where each finding has `toolSource` set (`"real-tool"` if the named tool actually ran and covered that finding, `"llm-reasoned"` with an explicit `fallbackReason` otherwise). Write its output to a scratch JSON file. 4. Dispatch the `assertion-auditor` agent AGAIN, in **Verify mode**, o