Mcp server design
Skill batteryshark/skill-tap/skills/development/mcp/mcp-server-design
Design FastMCP 3 tools, providers, transforms, authorization, background work, output contracts, and human approval boundaries for a production server.From its SKILL.md
npx -y skills add batteryshark/skill-tap --skill mcp-server-designAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
1.7 KB, 292 tokens by cl100k_base, as published. Nobody here has run it
Design a FastMCP Server
- Inventory users, component count, data sources, trust boundaries, latency, long-running work, and irreversible or external effects.
- Run
bin/mcp-server-design PATHto inventory existing FastMCP concerns and read references/architecture-decisions.md. - Keep domain logic library-first. Use providers to source components, transforms to shape the exposed catalog, and middleware for request execution concerns.
- Prefer ordinary directly exposed tools for a small catalog. Use Tool Search for on-demand discovery, and experimental Code Mode only when server-side Python composition justifies its sandbox and extra dependency.
- Make output match consumers: typed structured data for programmatic composition, concise content for model reading, and Apps when an interactive UI materially helps. Do not impose a universal text-only rule.
- Put authorization and visibility on components, authenticate HTTP deployments, keep secrets server-side, and require preview/approval for irreversible, financial, external-communication, permission, or sensitive-data actions.
- Use task support for genuinely long-running work and explicit timeouts for bounded calls. Keep interactive OAuth and unreliable network work out of startup.
- Verify schemas, catalog exposure, permissions, output shape, task behavior, and side-effect gates.
Use agents/architecture-reviewer.md for a threat- and token-budget-aware review.
What ships with it: 4 files
3.6 KB alongside SKILL.md, 2 of them executable
agents/
bin/
- mcp-server-designruns246 B
references/
scripts/
- collect_evidence.pyruns1.2 KB