audit-docs

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the project's codebase and documentation files, which could contain malicious instructions. * Ingestion points: Project files, including Markdown documentation, source code, and in-code comments, are read in full to verify claims (SKILL.md, Rule 2). * Boundary markers: While the skill mandates strictly factual output and grounding, it does not specify technical delimiters or escaping mechanisms to isolate content from instructions during data ingestion. * Capability inventory: The agent can read any file in the accessible workspace and write to documentation files. It also delegates tasks to sub-agents (agents/claim-verifier.md, etc.) which share these capabilities. * Sanitization: The instructions include a specific rule to redact credentials (e.g., API keys, tokens) found in the code, which reduces data exposure risk, but there is no explicit defense against malicious prompt instructions embedded in the processed files.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 04:37 AM