heal-skill
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill derives proposed code and instruction changes from the conversation history and error logs. If these inputs contain malicious instructions, the agent might suggest inserting them into other skills.\n
- Ingestion points: Conversation history, task context, and user-provided arguments in SKILL.md.\n
- Boundary markers: The skill uses a structured proposal format but does not explicitly warn the model to ignore instructions within the analyzed data.\n
- Capability inventory: The skill has the ability to Read and Edit files and perform git operations via Bash within the ./skills/ directory.\n
- Sanitization: No automated sanitization is present; however, the skill requires explicit user approval via AskUserQuestion before any edits are applied, which serves as a critical security checkpoint.
Audit Metadata