learning-playbook
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from project work and research, which presents a surface for indirect instructions.
- Ingestion points: Untrusted data enters the agent context through 'project work, research, or repeated agent friction' as specified in SKILL.md.
- Boundary markers: The instructions mandate establishing an 'evidence boundary' and identifying 'one counterexample or explicit boundary where the lesson does not apply'.
- Capability inventory: The skill has the capability to 'apply explicitly approved changes' to repositories, knowledge bases, and agent configurations.
- Sanitization: Risks are mitigated by the requirement for 'explicitly approved changes' and 'independent behavioral tests' before any guidance is promoted.
- [SAFE]: The skill consists exclusively of markdown instructions and documentation without executable scripts, binary files, or external network dependencies.
Audit Metadata