The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for ingesting and processing untrusted data from network APIs, HTTP requests, and file streams, which could be exploited via malicious data payloads.
Ingestion points: Processes data from fetch() responses (§A3, §E1) and HTTP request components like request.body and request.query (§B3, §D1).
Boundary markers: The skill recommends using TypeBox for strict JSON schema validation and structured verification of JWT fields.
Capability inventory: Includes capabilities for writing to the file system via pipeline() and making arbitrary network requests via fetch().
Sanitization: Encourages the use of @sinclair/typebox for schema validation and manual verification of security-critical JWT claims (exp, iss, aud).
[DYNAMIC_EXECUTION]: The skill provides boilerplate code and configuration for creating Node.js native C++ addons, facilitating the execution of compiled code within the Node.js runtime.
Evidence: Section C3 contains source code for addon.cpp and a binding.gyp configuration file to support the compilation and loading of native modules via N-API.
[EXTERNAL_DOWNLOADS]: The skill references and recommends the installation of several well-known libraries from the npm registry.
Evidence: Mentions standard ecosystem packages including fastify, @fastify/create-error, async-cache-dedupe, lru-cache, close-with-grace, @fastify/env, fastify-plugin, @sinclair/typebox, @fastify/jwt, @fastify/cors, @fastify/helmet, pino-pretty, @fastify/oauth2, why-is-node-running, and node-addon-api.

nodejs-development