The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and enrich external data ('leads' or 'rows'), which serves as an ingestion point for untrusted content. Malicious data could theoretically contain instructions intended to influence the agent's behavior.
Ingestion points: lead parameter in enrichOrFallback and reliableEnrich functions (SKILL.md); rows list in submit_rows and process_batch (references/implementation-guide.md).
Boundary markers: Absent; no delimiters or explicit instructions are provided to the model to ignore potential injection within the lead data.
Capability inventory: The skill allows Write, Edit, and Bash(curl:*) operations across its implementation files.
Sanitization: The provided reliability patterns focus on flow control and persistence, without explicit validation or escaping of the lead data content.
[COMMAND_EXECUTION]: The troubleshooting section in references/implementation.md provides shell commands for network diagnostics and API health checks.
Evidence: Includes curl for testing the Clay API and verifying connectivity, and dig for DNS resolution checks.
[EXTERNAL_DOWNLOADS]: The implementation guides reference standard external libraries and services.
Evidence: The Node.js implementation uses the opossum circuit breaker library, and the Python implementation uses the redis and requests packages. The skill interacts with the official API at api.clay.com, which is a well-known service.

clay-reliability-patterns