reflect

Pass

Audited by Gen Agent Trust Hub on Nov 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill analyzes conversation transcripts to suggest edits to agent skills, creating a surface for indirect prompt injection from past user inputs.
  • Ingestion points: Reads active conversation history via host APIs or the /recall command (SKILL.md).
  • Boundary markers: Explicitly instructs the agent to treat the transcript as data and to follow no instructions found within it (SKILL.md).
  • Capability inventory: The skill has the ability to propose and execute edits to skill source files using the /writing-for-agents tool (SKILL.md).
  • Sanitization: Requires explicit user approval before any skill is edited, providing a human-in-the-loop safeguard (SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 17, 2026, 09:14 AM