The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructs the agent to process potentially untrusted external data, such as UI content via Playwright and project files during validation, creating a surface for indirect prompt injection.
Ingestion points: Project files, UI content (Gate 2), and validation logs (Gate 1).
Boundary markers: None explicitly defined to isolate processed data from agent instructions.
Capability inventory: File system operations (deleting scratch files), process management (kill), and package execution (pnpx).
Sanitization: No explicit sanitization or validation of the processed data is described.
[COMMAND_EXECUTION]: The skill directs the agent to use powerful shell commands for environment cleanup and process management.
Evidence: Use of kill and kill -9 to terminate test servers, following a verification process using lsof and ps to identify the correct PID.
Evidence: Use of rm -rf to clear specific cache directories such as node_modules/.vite and node_modules/.cache/storybook to resolve test flakiness.
[EXTERNAL_DOWNLOADS]: The skill utilizes a package runner to execute a specific linting tool from a public registry.
Evidence: Execution of pnpx node-actionlint, which downloads and runs the node-actionlint package from the NPM registry.

done