The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download code from an external GitHub repository (github.com/jamiepine/voicebox.git) in references/generative-tools.md for voice synthesis functionality.
[REMOTE_CODE_EXECUTION]: In references/generative-tools.md, the skill directs the user to run npx create-video@latest. This command downloads and executes the latest version of a package from the NPM registry at runtime, which can be a vector for supply chain attacks if the package or its dependencies are compromised.
[COMMAND_EXECUTION]: The skill includes several shell command examples for interfacing with external APIs and local scripts:
curl commands in references/generative-tools.md for interacting with Google Gemini and ElevenLabs APIs.
Node.js CLI execution in SKILL.md (e.g., node tools/clis/google-ads.js) to manage ad campaign data.
ffmpeg execution for media processing.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it is designed to ingest and analyze external "performance data" from CSVs, pastes, or API outputs to generate new ad variations.
Ingestion points: External data provided by the user (CSV, paste, API output) as described in the "Iterating from Performance Data" section of SKILL.md.
Boundary markers: No specific delimiters or instructions for the agent to ignore embedded commands within the ingested data are present.
Capability inventory: The agent has access to command execution (via local Node scripts) and network operations (via curl).
Sanitization: There is no evidence of sanitization or validation of the input performance data before it is processed by the agent.

ad-creative