The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install and use third-party tools from the NPM registry. Specifically, it references npx react-doctor@latest scan in RUNTIME-TRACE.md and npm install -g @lhci/cli in LIGHTHOUSE-CI.md. These are well-known developer tools from a trusted public registry.
[COMMAND_EXECUTION]: The skill documents the use of various CLI tools for performance-related tasks, such as ffmpeg and cwebp for media optimization in BROWSER.md, and lhci for continuous integration in LIGHTHOUSE-CI.md.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external performance data, such as Lighthouse reports and runtime traces. This represents a potential attack surface where malicious data could influence agent behavior.
Ingestion points: Performance trace files (.json.gz) in RUNTIME-TRACE.md and Lighthouse JSON reports in LIGHTHOUSE-CI.md.
Boundary markers: No specific boundary markers or 'ignore' instructions are provided for the data processing steps.
Capability inventory: The skill utilizes shell command execution (via npx, npm, ffmpeg, cwebp, lhci) across multiple scripts.
Sanitization: No explicit sanitization or validation of the ingested JSON/trace data is mentioned in the instructions.

web-performance