The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from local repository files and GitHub API responses. Ingestion points include the inspect_repository.py script (which reads local project files) and gh api calls (which fetch remote metadata). The skill includes a 'Trust boundary' section in SKILL.md that serves as a boundary marker, explicitly directing the agent to treat files and API responses as data to inspect rather than instructions. Capabilities available to the skill include file writes and subprocess calls for scaffolding tools (e.g., uv, npm, xcodebuild). Sanitization is implemented in the Python script via regex validation of repository identifiers.
[COMMAND_EXECUTION]: The skill performs repository inspection using a bundled Python script and scaffolds projects using official generators such as create-expo-app, uv, and xcodebuild. These commands are scoped to the primary purpose of initializing and verifying new repositories.
[EXTERNAL_DOWNLOADS]: The skill fetches repository settings and metadata from GitHub's official API using the gh CLI. These operations target a well-known and trusted service for the purpose of verifying repository state.

bootstrap-repository