reflection

Pass

Audited by Gen Agent Trust Hub on Feb 25, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it ingests and analyzes user-controlled conversation history to propose configuration changes.\n
  • Ingestion points: Interaction history and tool failure logs are reviewed for analysis.\n
  • Boundary markers: No specific delimiters are used to separate user data from analysis instructions.\n
  • Capability inventory: The skill can propose modifications to other skill definitions and the CLAUDE.md environment file.\n
  • Sanitization: The skill includes a mandatory confirmation step where the user must review and approve any changes before they are applied, mitigating the risk of unintended behavior modifications.\n- [NO_CODE]: The skill consists exclusively of markdown instructions and metadata, with no accompanying scripts or executable files.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 25, 2026, 03:10 AM