The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted product descriptions and usage instructions from a local aeo.config.json file and interpolates them directly into generated files like agent-instructions.txt and llms.txt without sanitization. These files are intended to be served to external AI agents to influence their behavior and citations.
Ingestion points: aeo.config.json fields such as product.answer, aeo.whenToUse, and aeo.whenNotToUse in scripts/scaffold.mjs.
Boundary markers: The generated files use standard markdown formatting (headers, blockquotes) but do not include explicit instructions for agents to ignore malicious content within those sections.
Capability inventory: The scaffold.mjs script has local file-write capabilities (fs.writeFileSync).
Sanitization: No escaping or filtering is applied to the user-supplied strings before they are written to the output files.
[COMMAND_EXECUTION]: The instructions guide the user to run local Node.js scripts (scripts/scaffold.mjs and scripts/verify.mjs) to perform project modifications and network-based validation.
[DYNAMIC_EXECUTION]: The scaffold.mjs script acts as a code generator, creating multiple TypeScript and JavaScript files from hardcoded templates. While the templates are static, the output incorporates variable user configuration and modifies the local project structure.
[EXTERNAL_DOWNLOADS]: The verify.mjs script uses the fetch API to make network requests to a user-provided URL (e.g., localhost:3000 or a production domain) to verify the status and content of implementation endpoints.

agent-ready-nextjs