amend-skill

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from execution logs to determine how to modify other skills.
  • Ingestion points: Reads observations/<skill-name>/runs.md and skills/<bucket>/<category>/<skill-name>/SKILL.md. The run logs (runs.md) are likely to contain data, payloads, or responses from external targets audited by other skills.
  • Boundary markers: The instructions do not define specific delimiters or instructions for the agent to ignore potentially malicious content within the analyzed logs.
  • Capability inventory: The skill has the capability to read and write to the local file system (specifically skills/ and observations/ directories) to apply amendments.
  • Sanitization: There is no automated sanitization or validation of the log content; however, the skill incorporates a critical safety control by requiring explicit user confirmation before any proposed amendment is applied to a skill file.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 11:59 AM