update-project-skills

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Automates project maintenance by executing shell commands via npx and a local Python script to update, add, and remove skill definitions.\n- [EXTERNAL_DOWNLOADS]: Fetches and runs the latest version of the skills utility from the npm registry and references content from the toy-crane/skills GitHub repository.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests configuration data from skills-lock.json and agent manifests.\n
  • Ingestion points: skills-lock.json, companion-agents/manifest.json, and agent definition files (Markdown, TOML).\n
  • Boundary markers: The sync_companion_agents.py script serves as a validation layer, ensuring data is correctly formed and project-local before processing.\n
  • Capability inventory: Shell command execution (npx, python3) and atomic file system write access.\n
  • Sanitization: Implements path traversal protection through Path.resolve().relative_to(project) and uses the NAME_PATTERN regex to validate agent identities.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 04:26 PM
Security Audit — agent-trust-hub — update-project-skills