add-skill

Pass

Audited by Gen Agent Trust Hub on Jul 21, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill demonstrates an indirect prompt injection surface because it processes external data to generate or modify skill instructions.
  • Ingestion points: Processes user-supplied $ARGUMENTS, performs recursive directory reads of the .local/ path, and integrates results from web-search operations.
  • Boundary markers: The workflow does not explicitly define delimiters to isolate untrusted content within the generated SKILL.md files.
  • Capability inventory: Has the ability to create and modify files in the skills/ directory and execute shell commands via make check.
  • Sanitization: Includes instructions to clean facts before storage and mandates a manual review/owner approval step before committing changes.
  • [COMMAND_EXECUTION]: Executes the local make check command to validate the syntax and behavioral requirements of the skill being authored or modified.
  • [EXTERNAL_DOWNLOADS]: Performs web searches to identify current facts, official documentation, and primary sources used as reference material for skill creation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 21, 2026, 09:55 PM
Security Audit — agent-trust-hub — add-skill