The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests configuration data from environment variables, creating a potential surface for indirect prompt injection.\n
Ingestion points: Accesses environment variables such as process.env.NODE_ENV and CLAY_WEBHOOK_URL within the configuration logic in SKILL.md and references/implementation-guide.md.\n
Boundary markers: No explicit delimiters are used to isolate configuration inputs from the agent instruction context.\n
Capability inventory: The skill supports network operations through the Bash tool using curl.\n
Sanitization: Implements strict schema validation using the Zod library to ensure configuration integrity and filter malformed inputs.\n- [DATA_EXFILTRATION]: Describes diagnostic procedures that transmit authentication tokens to external service endpoints.\n
Evidence: Documentation in references/implementation.md provides shell commands utilizing curl to transmit the CLAY_API_KEY to the official api.clay.com endpoint.\n
Context: The network operations target the official API of a well-known service as part of standard setup verification.\n- [COMMAND_EXECUTION]: Provides guidance on executing shell commands for secret management on cloud platforms.\n
Evidence: Contains examples for using the GitHub CLI, AWS CLI, and Google Cloud SDK to store and manage per-environment secrets.

clay-multi-env-setup