The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external report data from Lighthouse, PageSpeed Insights, and the Chrome UX Report (CrUX) via scripts/triage-vitals.py and scripts/check-web-perf-facts.py.\n
Ingestion points: scripts/triage-vitals.py accepts untrusted JSON reports as a CLI argument or via stdin. scripts/check-web-perf-facts.py ingests network responses from external registries.\n
Boundary markers: The scripts do not implement specific boundary markers or "ignore" instructions for the ingested JSON content, relying on standard JSON parsing.\n
Capability inventory: The skill is allowed tool access for Bash, Read, and Write operations, which are used by the scripts to read files and output findings.\n
Sanitization: Ingested data is parsed as JSON but not explicitly sanitized, as the tool's purpose is to report on the values found within these technical reports.\n- [EXTERNAL_DOWNLOADS]: The scripts/check-web-perf-facts.py script performs network operations to ensure performance thresholds and tool versions are current.\n
Fetches package information from the official NPM registry (registry.npmjs.org).\n
Retrieves library constants from the Unpkg CDN (unpkg.com) for the web-vitals package.\n
These operations target well-known, industry-standard services for legitimate validation and maintenance purposes.

web-perf-ops