skill-improve

Pass

Audited by Gen Agent Trust Hub on Jun 27, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill derives updates from execution experience, which can include untrusted data from external sources or user interactions processed in previous tasks. This creates a surface where malicious input could influence the modification of skill instructions or scripts.
  • Ingestion points: Reads task context and session history.
  • Boundary markers: None specified for the input data.
  • Capability inventory: Edit and Write tools are used to modify SKILL.md and associated scripts.
  • Sanitization: Includes a Diff Preview and requires explicit user confirmation before any file modifications are performed.
  • [COMMAND_EXECUTION]: The skill is authorized to modify executable files within the resources of other skills, allowing for changes to script logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 27, 2026, 07:09 AM
Security Audit — agent-trust-hub — skill-improve