learn
Pass
Audited by Gen Agent Trust Hub on Apr 1, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it transforms arbitrary conversation content into persistent behavioral rules.
- Ingestion points: The skill processes the conversation history (SKILL.md) to identify mistakes or corrections.
- Boundary markers: Absent. The instructions do not define clear boundaries between trusted user instructions and untrusted data that may have been ingested during the session.
- Capability inventory: The skill allows for the modification of
./CLAUDE.mdand~/.claude/CLAUDE.md, which serve as persistent configuration files for the agent. - Sanitization: Absent. There is no requirement for the agent to validate that the identified lesson is safe or originates from a trusted intent before codifying it.
Audit Metadata