The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute Node.js scripts (snapshot.mjs, agent-readiness.mjs) and the Lighthouse CLI for page auditing.
[EXTERNAL_DOWNLOADS]: The skill invokes npx lighthouse, which fetches content from external URLs for auditing purposes and may download package updates from official registries. The documentation notes that this is a consent-gated operation requiring user agreement.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from web page snapshots to identify SEO and accessibility issues.
Ingestion points: The skill reads pages/<slug>.json and pages/<slug>.html files containing external web content (SKILL.md).
Boundary markers: Findings are required to conform to schema/finding.schema.json, providing a structured output format that limits the influence of processed data.
Capability inventory: The skill uses Read, Grep, and Bash (to run node and npx).
Sanitization: The instructions direct the agent to quote observed elements verbatim as evidence within the JSON schema, which helps isolate untrusted content from the agent's logic.

seo-agent-readiness