The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions facilitate the installation of the agent-browser utility from Vercel Labs' official GitHub repository to enable research capabilities.
[COMMAND_EXECUTION]: The skill utilizes shell commands for environment setup and tool discovery, specifically referencing scripts located within the .pawbytes project structure and common package manager commands like npx.
[DATA_EXFILTRATION]: The documentation provides guidance on using persistent browser profiles (e.g., ~/.linkedin-profile) for authenticated research tasks. Notably, it includes explicit security warnings to ensure sensitive session state files are excluded from version control using .gitignore.
[PROMPT_INJECTION]: The skill's research workflow involves an indirect prompt injection surface by ingesting untrusted content from external websites (Google, Reddit, Quora) via browser automation.
Ingestion points: Web content retrieved via agent-browser get text body in references/content-research.md.
Boundary markers: None explicitly defined in the research prompts to isolate ingested content.
Capability inventory: Browser automation, file system access for writing deliverables, and local command execution.
Sanitization: No automated sanitization or filtering of the external content is described before it enters the agent context.

paw-mkt-content