he-learn
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [NO_CODE]: The skill consists entirely of markdown instructions and process descriptions without executable code scripts or tool configurations.
- [INDIRECT_PROMPT_INJECTION]: The skill defines a process for ingesting external data such as task attempts and results to generate architectural decision records (ADRs) and new skill files. This creates a theoretical attack surface for indirect prompt injection from processed data.
- Ingestion points: The skill uses 'current task + observed attempts/results' as primary input in SKILL.md and references 'source evidence' in decisions.md.
- Boundary markers: The skill relies on structured templates (ADR and Skill formats) to organize ingested information and minimize confusion.
- Capability inventory: The agent is directed to write files to the local file system, specifically in 'docs/adr/' and '.agents/skills/'.
- Sanitization: The skill explicitly mandates 'Never copy secrets or raw transcripts,' which is a key security practice to prevent data exposure during the documentation process.
Audit Metadata