context-health-monitor
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill captures the agent's current state, including 'Files Involved' and 'What Was Attempted', and writes it to persistent files ('.gsd/STATE.md', '.gsd/DECISIONS.md') intended for later retrieval. This creates an attack surface where malicious instructions in processed data could be mirrored into the state dump and executed in subsequent sessions.
- Ingestion points: Data is pulled from the active agent session context in SKILL.md.
- Boundary markers: The provided markdown template for state dumps lacks delimiters or specific instructions to treat recorded context as inert data.
- Capability inventory: The skill utilizes file writing capabilities to create and update the '.gsd/' directory contents.
- Sanitization: No sanitization or escaping is performed on the context data before it is recorded to the state files.
Audit Metadata