The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a data ingestion surface where external lead data (names, domains) is processed. As this is the primary purpose of the lead enrichment workflow, the risk is inherent to the use-case.
Ingestion points: The enrichLeadList function in src/workflows/enrich-leads.ts accepts an array of LeadInput objects.
Boundary markers: The code snippet includes validation logic (l.domain?.includes('.')) and deduplication to ensure data quality, though it does not implement specific prompt-injection boundary markers as it targets a data API rather than an LLM prompt.
Capability inventory: The skill uses the clay.sendBatch method to transmit data to the Clay platform and describes the configuration of HTTP API calls for CRM export.
Sanitization: Basic validation is performed on input domains and names before processing.
[EXTERNAL_DOWNLOADS]: The skill provides links to official documentation at university.clay.com. These references are to well-known service domains associated with the skill's primary functionality.
[CREDENTIALS_UNSAFE]: The skill includes an example HTTP API configuration that uses the placeholder {{HubSpot API Key}}. This is a standard practice for documenting credential usage without exposing actual secrets.

clay-core-workflow-a