The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill architecture involves ingesting data from potentially untrusted external sources, which constitutes an attack surface for indirect prompt injection.\n- Ingestion points: The instructions specify reading from user-provided context, public data sources, scraped page text, reviews, comments, and third-party exports as seen in SKILL.md and references/skill-contract.md.\n- Boundary markers: The skill includes explicit instructions in SKILL.md and references/skill-contract.md to treat scraped material as untrusted input and directs the agent not to follow instructions embedded within that evidence.\n- Capability inventory: The skill facilitates interaction with various image generation models and marketing integration layers such as HubSpot, Salesforce, and Stripe as described in references/connectors.md.\n- Sanitization: The skill relies on natural language instructions to sanitize behavior when processing external data, rather than programmatic validation.

image