learner

Pass

Audited by Gen Agent Trust Hub on Apr 24, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to persist data by writing markdown files to specific directory paths such as .omc/skills/ and ~/.claude/skills/. This is a routine operation for a configuration management skill.
  • [INDIRECT_PROMPT_INJECTION]: The skill creates a surface where untrusted data from a conversation can be converted into persistent instructions for future agent sessions.
  • Ingestion points: Conversation history, including error messages and debug output (SKILL.md, Step 1).
  • Boundary markers: The skill requires the use of YAML frontmatter templates for the generated files but does not explicitly specify sanitization for the body content.
  • Capability inventory: File writing capabilities to project and user configuration directories.
  • Sanitization: Step 2 includes quality validation checks focused on utility and specificity, although it does not explicitly screen for malicious instructional content.
  • [DYNAMIC_EXECUTION]: The skill implements logic to generate and store new instructions (skills) using a predefined markdown template. This is consistent with the skill's stated purpose of self-improvement and learning.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 24, 2026, 08:49 AM