product-principles

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to store and load user preferences, decisions, and 'Taste entries' from project files to minimize human oversight in future runs. This pattern creates a surface where untrusted content within the project repository could influence the agent's behavior.
  • Ingestion points: Data is loaded from 'principles line', 'Taste entry', 'decision record', and 'project's own files' as described in SKILL.md.
  • Boundary markers: The principles do not specify the use of delimiters or 'ignore instructions' warnings when processing these files.
  • Capability inventory: The instructions imply the agent has capabilities to read and write to the project's file system and adjust its behavioral defaults based on that content.
  • Sanitization: No validation or sanitization of the content loaded from files is mentioned.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 10:34 PM