What problem does it solve? Security findings often stay theoretical: a scanner flags an issue, but nobody proves it is exploitable or verifies the fix actually closes it. This Skill embeds penetration testing into the development cycle, turning each proof of concept into a regression test that proves the patch works. ## Core Features & Use Cases - Scoped authorization gate: Requires a .pentest-scope.json file declaring targets, out-of-scope hosts, rules, and expiration before any offensive skill runs; production and third-party assets are always excluded. - Closed exploit-patch loop: For each vector it runs recon, builds a reproducible PoC, verifies the finding with a security reviewer, converts the PoC into a failing regression test via TDD, applies the fix, and re-runs the exploit to confirm it now fails. - Local reporting: Accumulates each closed vector (severity, PoC, patch, regression test, re-exploitation result) into a local report that never leaves the machine. - Use Case: While developing an API, you suspect an IDOR on an endpoint. The Skill checks your scope file, exploits the endpoint against your local Supabase instance, confirms the finding, writes a failing test, patches the shared authorization function, and re-runs the attack to prove it is dead. ## Quick Start Ask the AI to run /pentest against the authorized targets in your .pentest-scope.json and close one vulnerability vector end to end.
