The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could contain malicious instructions.
Ingestion points: Reads client-provided materials from the ~/marketing-plans/{client-slug}/materials/ directory and fetches data from external APIs (Ahrefs, Stripe, GA4, Customer.io).
Boundary markers: The instructions do not explicitly mandate the use of delimiters or 'ignore instructions' warnings when interpolating external content into the agent's context.
Capability inventory: The skill has capabilities to write files to the local system (final_plan.md, research.md), perform network operations via multiple MCPs, and utilize browser automation through agent-browser.
Sanitization: There is a 'verification pass' mentioned in the finalization phase to check for brand voice and machine-specific paths, but no specific sanitization or escaping of external inputs is defined.
[EXTERNAL_DOWNLOADS]: The skill instructions and documentation reference several well-known third-party services and CLI tools.
Services: Fetches data from Ahrefs, DataForSEO, Stripe, Google Ads, and Customer.io.
Utilities: References the use of agent-browser and defuddle for research and extraction tasks. These are well-known utilities for agent environments and do not trigger a verdict escalation under trusted source rules.

marketing-plan