The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external CRM systems, SQL databases, and email threads to automate lead tiering and outreach, which presents a surface for indirect prompt injection attacks.
Ingestion points: Data is retrieved from HubSpot (hubspot_search_contacts), a SQL database (execute_sql), and Gmail messages via MCP tools.
Boundary markers: The skill enforces strict procedural gates, specifically qualify_lead (Gate Q) for deduplication and exclusion rules, and check_my_copy (Gate B) for output validation.
Capability inventory: The skill can create Gmail drafts, update HubSpot records, and remove contacts from Apollo sequences.
Sanitization: All outbound strings are mandated to pass through the check_my_copy gate, and product claims are cross-referenced via search_product_catalog (Gate S).
[SAFE]: The skill architecture adheres to high safety standards, incorporating mandatory human-in-the-loop steps (Gmail drafts require approval before sequence enrollment) and robust self-healing logic. All external integrations involve well-known, trusted business services like HubSpot and Apollo. Telemetry and analytics are stored in platform-standard local directories (~/.claude/).

sales-revenue-skill