The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses exec() and compile() in scripts/jev_lib.py and scripts/test_offline.py to execute generated Python code locally. This provides a local preview of the lead-scoring logic to ensure parity with the cloud-based Clay workflow.
[COMMAND_EXECUTION]: The clay CLI and git command are executed via subprocess.run() in several scripts to manage workflows and perform security audits of local configuration files.
[INDIRECT_PROMPT_INJECTION]: The skill ingests contact data from Clay tables and evaluations are performed by the Jev model. Instructions use backticks as boundary markers to delimit data. The skill possesses capabilities for network operations and command execution. Inputs are truncated to a maximum character length as a sanitization measure.
[EXTERNAL_DOWNLOADS]: The skill communicates with TypeSafe (api.typesafe.ai) and OpenRouter (openrouter.ai) APIs for scoring inference. These network operations are standard for the skill's intended operation and target well-known service providers.

score-contacts-with-jev