The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The reference documentation in references/generative-tools.md provides instructions to clone an external repository from an untrusted GitHub user (jamiepine/voicebox).
[REMOTE_CODE_EXECUTION]: The skill recommends the use of npx create-video@latest, which facilitates the download and execution of code from the npm registry at runtime without integrity verification.
[COMMAND_EXECUTION]: The skill's workflow involves the execution of local management scripts such as node tools/clis/google-ads.js to fetch performance data and manage campaigns.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and iterate based on external performance data (CSV, pastes, or API outputs), creating a surface for indirect prompt injection if that data contains malicious instructions.
Ingestion points: Performance data is pulled from advertising platform tools or provided directly by the user as described in SKILL.md.
Boundary markers: The instructions lack explicit delimiters or warnings to the model to ignore embedded instructions within the performance metrics.
Capability inventory: The skill has the capability to perform actions on advertising platforms and execute local vendor scripts based on the processed data.
Sanitization: There are no specific validation or sanitization steps defined for the ingested data before it is analyzed by the agent.

ad-creative