skill-doctor
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPERSISTENCEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local bash scripts (orchestrate.sh and install-deps.sh) found in the plugin root to perform diagnostics, check authentication, and manage dependencies.
- [EXTERNAL_DOWNLOADS]: The skill recommends installing third-party tools such as rtk (via brew or cargo) and the agy (Antigravity) CLI, along with the @openai/codex package from a trusted organization.
- [PERSISTENCE]: The skill includes instructions to install a global bash hook via the rtk init -g command, which modifies the user's shell environment to persist across terminal sessions.
- [INDIRECT_PROMPT_INJECTION]: The skill reads project-level RUNTIME.md files to gather project-specific context, creating a potential attack surface. • Ingestion points: RUNTIME.md file in the project root. • Boundary markers: None identified. • Capability inventory: Shell script execution, package installation, and global hook modification. • Sanitization: No sanitization or validation of the RUNTIME.md content is identified.
- [DYNAMIC_EXECUTION]: The skill uses a bash resolver to dynamically search for the plugin root directory and executes scripts from the resulting path.
Audit Metadata