The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from project configuration files and incorporates it into a generated shell script without sanitization.
Ingestion points: The script scripts/setup-husky-coverage.py reads data from package.json (specifically the test script and dependencies), .husky-test-coverage.json, and various test runner configuration files (e.g., jest.config.json).
Boundary markers: There are no delimiters or explicit warnings to the agent to ignore instructions embedded within the ingested configuration files.
Capability inventory: The skill uses subprocess.run() to perform package installations and Path.write_text() to create executable scripts.
Sanitization: The value of the project's test script is interpolated directly into the .husky/pre-commit hook file, meaning a malicious script in package.json would be automatically included in the generated hook.
[COMMAND_EXECUTION]: The skill uses the Python subprocess module to manage environment setup and dependency installation.
Evidence: The run_command function in scripts/setup-husky-coverage.py executes commands such as npm install, bun add, and npx husky install. The implementation follows best practices by passing arguments as a list rather than a single shell string, mitigating direct command injection risks during the setup phase.
[PRIVILEGE_ESCALATION]: The skill modifies file system permissions to ensure the functionality of the generated git hooks.
Evidence: In scripts/setup-husky-coverage.py, the code calls pre_commit_path.chmod(0o755) on the newly created hook file. This elevates the file's permissions to allow execution, which is the intended behavior for a Husky hook but represents the creation of a new executable entry point in the repository.

husky-test-coverage