add-skill

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: Utilizes shell commands such as git and the GitHub CLI (gh) to automate local repository management, commit changes, and submit pull requests.
  • [EXTERNAL_DOWNLOADS]: Clones the official design-skills repository from the Automattic organization on GitHub if the directory is not found on the user's local system.
  • [PROMPT_INJECTION]: The skill ingests local file content or user-provided descriptions to generate new skill files. This is a standard functional requirement for a contribution tool, though it creates a surface for indirect prompt injection where content in the source file could theoretically influence the agent's behavior during the scaffolding process. Basic validation is performed on skill names and description length.
  • Ingestion points: Reads local files or text descriptions provided in the $ARGUMENTS variable.
  • Boundary markers: None present.
  • Capability inventory: Bash, Read, Write, Edit, Glob.
  • Sanitization: Validates that the skill name follows kebab-case and ensures the body is non-trivial.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 02:05 PM
Security Audit — agent-trust-hub — add-skill