The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to process untrusted web content, creating a surface for indirect prompt injection, though it includes mitigation markers.
Ingestion points: Content is ingested from external websites during authentication verification and snapshots in SKILL.md.
Boundary markers: The skill defines the [PAGE_CONTENT: ...] delimiter and provides explicit instructions for the agent to ignore instructions within these blocks.
Capability inventory: The skill can execute local Node.js scripts and manage authenticated browser sessions.
Sanitization: There is no automated sanitization of ingested content described; it relies on LLM adherence to the provided instructions.
[COMMAND_EXECUTION]: The skill executes a local script provided by the vendor to manage sessions and authentication.
Evidence: The skill invokes node /Users/avifen/.agentsys/plugins/web-ctl/scripts/web-ctl.js to perform its primary functions. This is a vendor-owned resource associated with the author 'avifenesh'.
[EXTERNAL_DOWNLOADS]: The skill references the installation of standard browser automation tools from well-known sources.
Evidence: The documentation suggests using npx playwright install chromium to resolve missing dependencies. Playwright is a well-known service provided by Microsoft.

web-auth