The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to verify factual claims by retrieving data from the internet using web_search and web_fetch as described in references/operating-principles.md. This introduces an attack surface where instructions hidden on external websites could attempt to influence the agent's behavior.
Ingestion points: External content retrieved via web_fetch and data from web_search results.
Boundary markers: The instructions do not specify the use of delimiters or clear boundaries to isolate ingested web content from the system's core operating instructions.
Capability inventory: The skill provides access to web search, URL fetching, and a suite of GTM-specific tools including epic_gtm_audit and epic_quick_score.
Sanitization: While the instructions emphasize manual verification by the agent, there are no programmatic sanitization or escaping protocols defined for the ingested content.

helix-mcp-gtm-alpha