self-growth
Pass
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill is designed to ingest untrusted input from external tasks (Step 1) and record it in long-term memory or generate new skills (Step 3). This creates a vulnerability where malicious instructions in task data can be persisted into the agent's permanent knowledge base.
- Ingestion points: Task descriptions or knowledge items processed through the Five-Step Flow in SKILL.md.
- Boundary markers: The skill does not define specific delimiters or instructions for the agent to ignore embedded commands within the ingested data.
- Capability inventory: The skill relies on the
memory_storetool and instructions for updating configuration files likeCLAUDE.mdand creating new files in.claude/skills/. - Sanitization: The skill lacks any requirements for sanitizing, escaping, or validating the content before it is stored in long-term memory or translated into new skill instructions.
Audit Metadata