The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands in Step 6 to verify the installation by executing a local Python hook script (session-start.py). It uses a find command to locate the script within the plugin's own directory structure if environment variables are not present.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it processes data from the local repository to guide its setup process.
Ingestion points: Reads repository root file listings, project directory names, and identifies existing documentation folders (e.g., docs/, documentation/) and environment files (e.g., .env, .env.local).
Boundary markers: The skill scaffolds configuration in structured formats (JSON for .claude/settings.json) and standard Markdown templates, though it does not explicitly define LLM boundary markers for the repo data it reads.
Capability inventory: The skill uses Read, Write, Edit, Glob, and Grep tools and executes local bash/python commands to verify setup.
Sanitization: The skill explicitly instructs the agent to use narrow allow rules for scripts rather than broad wildcards to prevent the execution of arbitrary written files.
[SAFE]: The skill implements protective security measures by default, such as proposing deny rules for sensitive environment files and .pem keys, and ask rules for destructive GitHub (gh) and Docker operations.

setup