AI can produce a plausible patch around an unsafe import, dependency, or exception path in seconds. Plausible is not verified. Patchward runs five real scanners on your Python repo and opens one bounded fix PR at a time — with a receipt that states exactly what it checked.
Static analysis runs entirely on-premise in a network-isolated sandbox. Findings are normalized, triaged, and handed to a fix generator that writes one targeted patch — validated before it ever reaches a pull request.
Free, self-hosted CLI. Install it yourself (`uv tool install patchward`) and run it with your own Anthropic API key and GitHub token — your code and credentials never pass through Patchward's own infrastructure. There is no hosted service, no paid tier, and no sales process. Gate 3 runs in the normal case. The disclosed-skip path exists for any environment without a detected test runner and is drawn with equal weight, not hidden.
Gate 3 (test suite): PASS — your test suite ran and passed. All three gates verified.
Gate 3 (test suite): SKIPPED — no test runner was available in this environment, so this fix was not validated against your tests. Gates 1 and 2 passed: the flagged rule no longer fires, and the change stays inside the authorised lines. Run your suite before merging.
Pulled live from the site's own canonical facts ledger — not hand-typed into this page. If it goes stale, it's wrong in exactly one place to fix.