The Agent Skills Directory

[COMMAND_EXECUTION]: The test script scripts/test_api_request.py uses subprocess.run to execute the local utility script api_request.py. While this is intended for automated testing, it demonstrates the capability to launch external processes from within the skill's file structure.
[DATA_EXFILTRATION]: The skill includes a custom HTTP client, scripts/api_request.py, which is capable of sending data to arbitrary external domains.
Evidence: The script uses urllib.request.urlopen to connect to any domain provided via the --base-url argument.
Safeguards: The script enforces HTTPS (except for localhost testing) and implements automated redaction of sensitive fields (e.g., 'token', 'api-key', 'secret') in logged outputs and URLs to prevent accidental credential leakage.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external marketing tools (e.g., CRMs, advertising platforms, and analytics providers), creating a surface for indirect prompt injection.
Ingestion points: External data enters the agent context through responses fetched by scripts/api_request.py from marketing API endpoints.
Boundary markers: The skill instructions lack explicit delimiters or instructions for the agent to treat API response content as untrusted data.
Capability inventory: The agent has network access via the provided script and is encouraged to perform state-changing mutations on external systems once authorized.
Sanitization: While the helper script redacts secrets from its own output, there is no evidence of sanitization or filtering of the retrieved content to prevent it from influencing the agent's behavior maliciously.

marketing-integrations