The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze external project files and performance report data, creating a potential surface for indirect prompt injection where instructions could be embedded in the audited data.
Ingestion points: The agent is directed to process project source code (HTML, CSS, JS) and performance audit metrics (Lighthouse, DevTools) as described in the Decision Tree in SKILL.md and the measurement workflows in references/core-web-vitals.md.
Boundary markers: Absent. The instructions do not include specific delimiters or warnings to ignore instructions that might be embedded within the content being audited.
Capability inventory: The skill describes the use of package managers (npm), build tool commands (npx), and network requests for production reporting (fetch('/analytics')), which could be leveraged if an injection occurred.
Sanitization: Absent. No steps are provided to sanitize, filter, or validate the content of the analyzed files.
[EXTERNAL_DOWNLOADS]: The skill references several external resources, tools, and libraries for performance measurement and optimization.
Recommends the installation of various Node.js packages for bundle analysis and image processing (e.g., sharp-cli, rollup-plugin-visualizer).
Integrates the treosh/lighthouse-ci-action GitHub Action for automated performance auditing.
Configures connections to established services like fonts.googleapis.com and fonts.gstatic.com for font delivery.
[COMMAND_EXECUTION]: The skill provides detailed instructions for using command-line tools to analyze and optimize project assets.
Documented commands include standard package management operations via npm install and the execution of development utilities using npx.
Provides shell commands for image format conversion and optimization using sharp-cli, cwebp, and convert (ImageMagick).

web-performance