The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands using node, curl, npx, and ffmpeg to retrieve ad performance metrics and render video creative.
[EXTERNAL_DOWNLOADS]: Fetches assets and references configuration from well-known AI service providers including Google Gemini, OpenAI, ElevenLabs, and Replicate. It also references the open-source Voicebox project for local audio generation.
[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection via the processing of untrusted external content. 1. Ingestion points: Ingests marketing context from local files (.agents/product-marketing-context.md) and ad performance data from platform-specific API outputs or manual pastes. 2. Boundary markers: Lacks explicit delimiters or instructions to prevent the agent from obeying commands embedded within the processed context or data. 3. Capability inventory: The agent has access to shell command execution (node, npx), file system access, and network operations. 4. Sanitization: There is no evidence of input validation or sanitization before external data is used in prompts or processing pipelines.

ad-creative