The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data during its 'Live Website URL Audit' and 'Existing Local Codebase' modes, which creates a surface for indirect prompt injection.
Ingestion points: Defined in references/shared-patterns.md, the agent is directed to audit live URLs and perform deep inspection of local project files.
Boundary markers: The instructions lack specific delimiters or negative constraints to prevent the agent from executing instructions that may be hidden within analyzed third-party content.
Capability inventory: The skill has access to browser automation (agent-browser), web retrieval tools (WebFetch), and the ability to write files to the local marketing suites directory.
Sanitization: No procedures are specified for sanitizing or validating external data before it is incorporated into the agent's reasoning context.
[EXTERNAL_DOWNLOADS]: The documentation in references/shared-patterns.md provides instructions for downloading and installing the agent-browser tool from Vercel Labs' official GitHub repository. Vercel is a well-known and trusted organization.
[COMMAND_EXECUTION]: The skill references the execution of setup and discovery scripts (tool-discovery.sh, chrome-profiles.sh) located in a related vendor skill directory (paw-mkt-setup). These are standard scripts used for checking environment requirements and identifying browser profiles for research tasks.

paw-mkt-referral