TaintLens Code Audit

Audit code the way an attacker reads it: entry points, source-to-sink taint paths, per-object authorization checks. A SKILL.md agent skill with a 7-pass method, 0-10 severity rubric, language sink cheat-sheets (Python/JS/TS/Java/Go/PHP), and an output contract demanding file:line evidence plus a minimal PoC for every finding. Forged in a production bug-bounty/SRC pipeline. Works with Claude Code, DeepSeek Harness (dsh), Cursor, Codex-style CLIs and any SKILL.md-compatible agent.