The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references a GitHub template repository atomicpages/mcp-template for scaffolding new projects. This is a vendor-owned resource used for legitimate development purposes.
[COMMAND_EXECUTION]: Instructions include standard development commands for project initialization and building, such as bun install, bun run build, and running the openapi-ts code generator.
[INDIRECT_PROMPT_INJECTION]: As the skill describes an architecture for servers that process external API data and user-provided tool parameters, it acknowledge the potential attack surface. It mitigates this by strongly recommending Zod-based request validation and providing specific logic to ensure write-operations require explicit user confirmation via the readOnlyHint flag.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill provides explicit guidance on avoiding data exposure, specifically warning against logging Authorization headers or raw secrets and recommending that all debug output be sent to stderr with a stable, grep-able prefix.
[DYNAMIC_EXECUTION]: Recommends the use of dynamic import() statements specifically for edge runtimes (like Cloudflare Workers) to manage CPU startup budgets during module evaluation. This is a standard performance optimization technique and does not involve executing untrusted strings.

mcp-openapi-typescript-stack