reflection
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes conversation history and tool outputs to propose updates to agent skills and configuration files (CLAUDE.md).
- Ingestion points: Processes conversation history and tool call logs (SKILL.md).
- Boundary markers: No explicit delimiters or warnings to ignore instructions embedded in the analyzed history are defined in the skill logic.
- Capability inventory: Capable of proposing and applying modifications to skill definitions and local configuration files that define agent behavior.
- Sanitization: No sanitization or filtering of the ingested conversation data is specified before it influences the reflection process.
Audit Metadata