The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the clay CLI to interact with the Clay platform. Commands include clay whoami for environment verification, clay search for candidate discovery, and clay workflows / clay routines for process orchestration.
[DYNAMIC_EXECUTION]: To handle data enrichment at scale, the skill dynamically generates and publishes Clay workflows at runtime using clay workflows create, clay workflows nodes create, and clay workflows publish. These workflows are then executed via clay routines runs start to process candidate batches.
[INDIRECT_PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests and processes untrusted external data from candidate profiles, including headlines, summaries, and job descriptions.
Ingestion points: Data enters the agent context via clay search query-mode run results and the cpj-enrich-person-from-search enrichment action.
Boundary markers: The instructions require the agent to quote the specific profile text used for scoring, providing a level of evidence-based verification.
Capability inventory: The skill can perform searches, create/publish/run workflows on the Clay platform, and generate pre-filled email compose links.
Sanitization: The skill uses URI encoding for the generated mailto: and Gmail compose links to ensure data is correctly formatted.
[DATA_EXPOSURE]: The skill processes sensitive candidate information, including names, employment history, and contact details (if requested). It generates mailto: and Gmail compose URLs containing these details. However, the skill explicitly designs these to land as drafts in the recruiter's mailbox, ensuring no data is sent to external parties without manual human intervention.

source-candidates