sliced-bread-depth

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill contains a vulnerability where malicious files can influence the agent's assessment by leveraging the instruction to trust documentation.
  • Ingestion points: The skill processes source code files throughout the repository, specifically targeting docstrings and code comments to determine 'intent'.
  • Boundary markers: The instructions lack delimiters for the external content and explicitly mandate honoring declarations of 'intent' found in the code, which functions as an instruction to ignore architectural metrics.
  • Capability inventory: The skill reads and classifies all files in the repository and performs logic-based assessment based on LOC and structure.
  • Sanitization: The skill lacks validation or filtering for the content of docstrings, allowing the 'documented intent' check to be used as a bypass mechanism.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 07:34 AM
Security Audit — agent-trust-hub — sliced-bread-depth