The Agent Skills Directory

[PROMPT_INJECTION]: The skill explicitly instructs the agent to bypass safety and legal filters concerning protected characteristics such as age and ethnicity. It directs the agent to provide a single warning and then proceed to "build what they asked for," stating that the skill "does not refuse" these requests.
[COMMAND_EXECUTION]: The instructions rely heavily on the clay CLI tool to interact with the user's workspace. It includes commands to check identity (clay whoami), search for field references (clay search query-mode reference), and list workspace routines.
[DYNAMIC_EXECUTION]: The skill provides a detailed procedure for the agent to dynamically construct a new "Candidate Profile Enrich" workflow at runtime. This involves creating nodes and triggers using clay workflows nodes create and clay workflows triggers create, and then publishing the workflow for routine execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text from external sources, specifically candidate profile headlines, summaries, and experience descriptions fetched via Clay enrichments. These ingestion points (Step 6) lack explicit sanitization or boundary markers, and the data is subsequently used for agent-based scoring and outreach drafting.
[EXTERNAL_DOWNLOADS]: The skill fetches external configuration data and schemas for specific packages from the Clay platform using the clay workflows actions schema command during the execution flow.

source-candidates-2