The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Zernio MCP ads_create_standalone_ad tool to perform live advertising actions on Meta's platform. While this involves financial commitments and content publication, the risk is mitigated by a mandatory human-in-the-loop approval step (Step 4 and Step 6) before any ads are launched or money is spent.
[EXTERNAL_DOWNLOADS]: The skill retrieves creative assets and campaign data from external sources including Google Drive (via search_files) and Airtable (via list_records_for_table). These downloads are integral to the skill's primary function and targets well-known services.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it parses filenames from Google Drive and records from Airtable to generate advertising copy.
Ingestion points: External filenames and Airtable record fields are processed in Step 3 and Step 4.
Boundary markers: The skill includes an explicit warning about untrusted third-party content and structured approval prompts.
Capability inventory: The agent can read files (Drive/Airtable) and execute advertising operations (Zernio MCP).
Sanitization: Manual human review and approval of all generated copy is required before the skill proceeds to the final execution phase, providing a strong control against malicious injection in the source data.

launch-meta-ads