codebase-knowledge-builder

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an inherent attack surface because it is designed to ingest and process untrusted data from external codebases.\n- Ingestion points: The skill reads files, comments, documentation, and issues from a target repository during the reconnaissance and deep-dive phases (SKILL.md, references/recon-checklist.md, references/deep-dive-methodology.md).\n- Boundary markers: The skill includes a 'Trust and Safety Contract' that mandates treating all repository content as 'untrusted evidence' and explicitly forbids it from expanding scope or overriding instructions (SKILL.md).\n- Capability inventory: The skill performs file system reads and writes. It explicitly prohibits executing target code, installing dependencies, or invoking build scripts (SKILL.md).\n- Sanitization: The instructions require redacting secrets and escaping target-controlled text in the output artifacts to prevent the creation of active Markdown or HTML structures (SKILL.md, references/deep-dive-methodology.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 10:52 AM