The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python's subprocess.run in scripts/build_loader.py and scripts/load_drive.py to interact with the clay command-line tool. This interaction is necessary for managing workspace resources such as Audiences fields and record-writing workflows. The commands are constructed using list-based argument passing, which prevents shell injection vulnerabilities by ensuring that user-provided mapping values or file data are not interpreted as commands.
[DYNAMIC_EXECUTION]: The scripts/test_loader.py script utilizes the compile() function to validate the syntax of a Python code template intended for a Clay 'code' node. The code being compiled is a hardcoded static string within the skill, and the operation is performed locally as part of a validation suite for the developer or installer. This does not involve the execution of untrusted or externally sourced code.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from CSV and Excel files provided by the user.
Ingestion points: Data ingestion occurs in loader_lib.py, which supports CSV and native Excel (.xlsx) parsing using only standard library components.
Boundary markers: The skill relies on structured file parsing and explicit data type definitions (e.g., number, email, URL) to ensure input data is treated as values rather than instructions.
Capability inventory: The skill possesses the capability to manage Clay workspace resources via the CLI and perform local file operations.
Sanitization: The skill performs extensive type-checking and validation using loader_lib.parses_as. It detects and reports malformed data (such as range strings in numeric fields or malformed URLs) to the user, preventing corrupted data from being loaded into the workspace.

tam-audience-loader