The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources to generate outreach content and update workspace state, creating a surface for indirect prompt injection where malicious data could influence agent behavior.
Ingestion points: Data enters the agent context via gtm-context.md, evidence-and-inputs.md, and the retrieval of external CRM records and call history (references/content-contracts.md).
Boundary markers: The instructions mandate the separation of copyable payloads from source notes and the distinct labeling of planning placeholders (references/content-contracts.md).
Capability inventory: The agent can generate emails, LinkedIn messages, and call preparation documents, and it possesses capabilities for workspace mutations and public hosting via publish manifests (SKILL.md).
Sanitization: The skill includes mandatory validation steps requiring the agent to check all engine results for context fidelity and promise-justification, with explicit instructions to repair violations before final delivery (references/content-contracts.md).

generate