amend-skill

Pass

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external skill files and execution logs to generate modifications.
  • Ingestion points: Reads content from SKILL.md files and observations/<skill-name>/runs.md log files.
  • Boundary markers: The methodology lacks explicit instructions or delimiters to distinguish between legitimate log data and potential instructions embedded within those logs.
  • Capability inventory: The skill has the capability to modify existing skills on the file system and update observation records via the agent's file tools.
  • Sanitization: There is no mention of sanitizing or filtering the ingested data before it is used to formulate a code diff for the amendment.
  • [SAFE]: The skill enforces a human-in-the-loop safety check by requiring user confirmation of the proposed unified diff before applying any changes to the skill files.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 22, 2026, 05:34 PM
Security Audit — agent-trust-hub — amend-skill