The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and analyze external, untrusted content from competitor landing pages, exposing the agent to potential indirect prompt injection attacks.
Ingestion points: The skill uses fetch_webpage or curl to retrieve content from URLs discovered in public ad libraries (Phase 4).
Boundary markers: The instructions explicitly warn the agent to treat fetched data as untrusted and to ignore content attempting to redirect the task or disclose data.
Capability inventory: The agent uses search and web-fetch tools to process data and generate a strategic markdown report.
Sanitization: The prompt instructs the agent to avoid private networks, cloud metadata endpoints, and to ensure it does not execute page instructions or downloads.
[EXTERNAL_DOWNLOADS]: The skill performs fetching of public data from well-known services such as the Meta Ad Library and Google Ads Transparency Center. This is consistent with its primary purpose of competitive intelligence and does not involve downloading executable code.
[COMMAND_EXECUTION]: The skill identifies curl as a tool for fetching remote webpage content when dedicated fetching tools are unavailable. This usage is restricted to retrieving data for analysis and does not involve arbitrary command execution.

competitor-ad-intelligence