The Agent Skills Directory

[Data Ingestion and Processing]: The skill is designed to analyze and edit external documentation and codebases, which involves processing potentially untrusted data. The instructions mitigate the risk of indirect prompt injection by explicitly directing the agent to treat documents as source material only and to ignore any instructions embedded within them.
[System Tool Usage]: To perform its technical writer duties, the skill utilizes standard system utilities such as rg (ripgrep) for searching implementation details and mktemp for managing temporary files. These tools are used within their typical functional scope for code analysis and internal task management.
[Agent Delegation Framework]: The instructions authorize the use of subagents for parallelized technical verification tasks. This capability is managed through specific boundary definitions, specifying that subagents should have clear questions and restricted file access, and requiring the main agent to verify all evidence returned.
[Technical Grounding]: The skill emphasizes grounding documentation in evidence from the repository, using targeted searches and inspection of public exports, types, and tests. This focus on verification against the local source-of-truth helps ensure the accuracy of the generated content.
[Vendor Resource Integration]: The skill references and integrates with official documentation sites such as ai-sdk.dev and nextjs.org. These resources are maintained by the vendor and provide a reliable context for technical writing and integration tasks.

technical-writer