The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires cloning a repository from a third-party GitHub account (https://github.com/pennydinh/marketing-pineline-share.git) and installing dependencies via npm, yarn, or pnpm.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by crawling external sources such as TechCrunch, Twitter, and LinkedIn and feeding that data into AI prompts.
Ingestion points: The autoResearch and generateCustomContent functions in SKILL.md ingest untrusted data from web and social media sources.
Boundary markers: The provided prompt examples (e.g., Research insights: ${JSON.stringify(researchData.insights)}) lack clear delimiters or specific instructions to the model to ignore potential commands embedded within the fetched content.
Capability inventory: The skill has capabilities for network operations (API calls), file output (video rendering), and command execution (via Remotion CLI), which could be misused if the model follows instructions hidden in crawled data.
Sanitization: There is no evidence of sanitization, filtering, or validation of the research data before it is interpolated into prompts.
[DYNAMIC_EXECUTION]: The skill uses @remotion/bundler and @remotion/renderer to bundle React-based templates and render video media at runtime. This process involves the dynamic assembly and execution of code based on project inputs.

ultimate-ai-content-pipeline