Applies a user-supplied corpus of secure-coding rules (semgrep packs, SARIF rule packs, markdown rule sheets, custom YAML, or the user's claude-secure-coding-rules-style repo) to a specific project. Detects the project's language, framework, and AI/ML stack; selects the subset of corpus rules that apply; flags rules whose preconditions are unmet; produces a per-finding report with file + line + rule id + recommended fix. Refuses to fabricate rules from memory when no corpus is supplied. Use when reviewing a Python+FastAPI+LangChain or similar AI/ML app against a known rule corpus, hardening a polyglot repo with partial rule coverage, or surfacing which corpus rules apply …
