reflect
Pass
Audited by Gen Agent Trust Hub on Feb 28, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE]: The skill is implemented entirely in markdown and metadata within the SKILL.md file. No scripts or executables are included.
- [SAFE]: Analysis identified no malicious patterns, data exfiltration, or obfuscation. The skill's primary function is to provide structured self-reflection for the agent.
- [PROMPT_INJECTION]: The skill's process of scanning conversation history for reflection creates an indirect prompt injection surface. This is mitigated by the fact that the skill only suggests changes as diffs and requires human confirmation before execution.
- Ingestion points: Conversation history.
- Boundary markers: Absent.
- Capability inventory: Proposing modifications to agent instructions.
- Sanitization: Absent; relies on user verification.
Audit Metadata