verifying-claims
Check that what a document says about code is true — by reading the document,
the code, and the tests together and reporting where they disagree.
The reviewer is the agent, not a parser. There is no claim-comment DSL: the
prose's meaning is read directly and compared to what the code does and what the
tests assert. No shadow copy, because the thing checked is the thing the human
reads.
See SKILL.md for the procedure, the verdicts (PASS / FAIL / UNSUPPORTED /
STALE), and the division of labor with TDD.
Quick start
python3 scripts/gather_context.py --doc README.md --src pkg/ --tests tests/
That bundles the document text, the public API surface (ast-parsed, never
imported), and the test inventory into one report. The agent then reads the
bundle and judges each prose claim against it.
What this is and isn't
- Is: a triggered, semantic review of whether documentation matches reality
— run before docs ship, after a refactor, or as a sweep. - Isn't: a CI merge gate or a test framework. The deterministic behavioral
gate is your test suite (TDD). This is the prose layer tests can't reach.
Files
scripts/gather_context.py— deterministic input bundler.references/drift-report-example.md— what a review report looks like.
Skill folder: verifying-claims
Release of verifying-claims version 0.2.0
📥 Download & Install
⬇️ Download verifying-claims.zip
To install:
- Click the download link above (ignore the "Source code" archives below - they're auto-generated by GitHub)
- Go to Claude.ai Skills Settings
- Upload the downloaded ZIP file
- Requires paid Claude Pro or Team account
See official documentation for more details.
Recent Changes
d93d7fb fix(verifying-claims): valid YAML frontmatter — colon-space in description broke parsing (#692)
2238e93 verifying-claims v0.2: pivot from claim-DSL to agent-driven doc/code/test review (#691)
d646568 docs: Update CHANGELOG.md for released skills