The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes subprocess and os.execvpe across several scripts (launch.py, doctor.py, stop.py) to manage the application's lifecycle. These scripts perform process ownership verification using system tools like lsof and ps to ensure the agent only interacts with processes it started.
[EXTERNAL_DOWNLOADS]: Documentation in references/npm-install.md describes standard procedures for verifying published packages by querying the official NPM registry (registry.npmjs.org) for versioning and integrity data.
[DYNAMIC_EXECUTION]: The test suite in tests/test_ownership.py programmatically generates and executes a temporary HTTP server script. This is used as a fixture to validate that the skill's safety checks correctly identify and reject unauthorized processes.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from local API endpoints (/api/health, /api/companies) and JSON configuration files to corroborate the success of UI journeys. These ingestion points are used for programmatic validation and state reporting.
[SAFE]: The instructions include explicit guidelines to avoid recording credentials or connection strings. It employs isolated temporary directories for data and evidence, adhering to security best practices for automated verification environments.

verify-paperclip