The Agent Skills Directory

[SAFE]: A thorough analysis of the skill's instructions, references, and evaluation cases revealed no malicious patterns, obfuscation, or unauthorized data access. The skill's operations are transparent and strictly limited to its stated purpose of ad creative generation and performance iteration.
[PROMPT_INJECTION]: The skill includes a workflow for iterating on creative based on user-provided performance data (CSV, pasted text, or API output). While this creates an indirect prompt injection surface for untrusted data, it is a standard feature for this type of marketing tool and is handled safely within the documented context. Ingestion points: Performance metrics and product marketing context files. Boundary markers: Not explicitly defined. Capability inventory: Use of local CLI tools and network requests for API interaction. Sanitization: Not specifically detailed in the prompt instructions.
[COMMAND_EXECUTION]: The skill documentation describes the use of local Node.js scripts (located in tools/clis/) and shell commands like curl and ffmpeg to manage ad campaigns and process creative assets. These commands are scoped to the intended functionality.
[EXTERNAL_DOWNLOADS]: The skill provides integration guidance and API examples for reputable third-party services including Google DeepMind, ElevenLabs, and Remotion. All referenced external resources are well-known technology providers.

ad-creative