The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by crawling and parsing external websites including HTML tags, robots.txt, and sitemaps.
Ingestion points: scripts/lib/pages.mjs (crawling logic).
Boundary markers: None present to distinguish external site data from instructions.
Capability inventory: Network access via fetch and Playwright; filesystem access via the agent's ability to modify project files to fix identity issues.
Sanitization: Basic HTML entity decoding and regex-based extraction of metadata fields.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to user-provided URLs to inspect site metadata and assets for audit purposes.
[COMMAND_EXECUTION]: The skill executes local Node.js scripts such as inspect-metadata.mjs to perform technical audits of the target website.
[DYNAMIC_EXECUTION]: The skill uses dynamic imports in scripts/lib/browser.mjs to optionally load the playwright library from the local environment for headless browsing support.

launch-identity