agent-instructions

Pass

Audited by Gen Agent Trust Hub on Sep 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of repository data and existing instruction files to generate updated instructions, establishing a surface for potential indirect prompt injection attacks. \n
  • Ingestion points: The skill reads various instruction files such as AGENTS.md and copilot-instructions.md from both project and user directories as outlined in the Inputs section of SKILL.md. \n
  • Boundary markers: A 'Discoverability Filter' and '3-Question Check' are described in SKILL.md to regulate content inclusion, but these are instructional guidelines rather than programmatic boundaries. \n
  • Capability inventory: The skill is capable of writing to and updating multiple instruction files on the local filesystem. \n
  • Sanitization: No programmatic sanitization or encoding of external data is implemented; validation is based on manual oversight and agent logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 28, 2026, 05:29 PM