The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests configuration data from project files, creating a vulnerability surface for instruction injection.
Ingestion points: CLAUDE.md (read via grep for preferences), Gemfile, package.json, Cargo.toml, requirements.txt, pyproject.toml, go.mod, and config/database.yml (read for adapter type).
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present when reading these files.
Capability inventory: The skill possesses extensive execution capabilities including package managers (npm, pip, bundle, cargo, go), Rails-specific commands, and local Ruby script execution. It can also modify the .gitignore file.
Sanitization: Content extracted from CLAUDE.md is interpolated into shell commands without explicit validation or escaping.
[DYNAMIC_EXECUTION]: The skill uses environment variables to construct and execute command paths.
Evidence: It executes a Ruby script via ruby "${CLAUDE_PLUGIN_ROOT}/skills/using-sqlite-worktrees/scripts/create_sqlite_worktree.rb", which constitutes dynamic loading from a computed path based on platform-provided environment variables.
[COMMAND_EXECUTION]: The skill performs automated build and test operations upon detecting specific manifest files.
Evidence: Execution of bundle install, bin/rails db:test:prepare, npm install, cargo build, and pip install -r requirements.txt occurs automatically after worktree creation.
[EXTERNAL_DOWNLOADS]: The skill facilitates the retrieval of external code via standard development tooling.
Evidence: Triggers remote dependency downloads through npm, pip, bundle, cargo, and go targeting their respective well-known public registries.

using-git-worktrees