fuzz

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). The fuzz workflow ingests repository source code from FUZZ_CORPUS_DIR / the “canonical corpora” sync scripts (cloning/syncing external repos), and that repo body text is then parsed into LLM-readable diagnostics/reproducers, so outsider-authored free-form code can reach the agent’s LLM context via the fuzz findings/reproducer text.