The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: Analyzed for potential indirect prompt injection as the skill ingests untrusted data from the repository being processed.
Ingestion points: Project files including package.json, node_modules/next/dist/docs/, and internal repository documentation are read during Step 0 and Step 1 of the skill workflow.
Boundary markers: The skill uses <EXTREMELY-IMPORTANT> tags and explicit "Success criteria" to bound the agent's behavior to the project's established architecture.
Capability inventory: The skill has access to Bash, Read, Write, Edit, and Grep tools to modify the filesystem and run project-specific scripts.
Sanitization: The skill instructions promote the use of DOMPurify and Zod for sanitizing data within the application context, minimizing the risk from untrusted inputs.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform standard development tasks such as running linters, generating types, and executing tests (e.g., next typegen, eslint, Playwright). These operations are necessary for its primary purpose of building and verifying Next.js code and are executed within the context of the project's own build toolchain.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions for upgrading framework dependencies like Next.js and React using standard package managers (npm, pnpm). These references target official and well-known registries, which is standard practice for software development and does not pose an atypical security risk.

nextjs