mcp-engineering
Skill1 starsChanged 42 days agoApache-2.0View on GitHub
Design, implement, secure, test, and deploy Model Context Protocol clients and servers, including tools, resources, prompts, authorization, and interoperability.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- bepaldes/heimdall-agenttheir own version, first seen Nov 8, 2026
- RuggG/mcp-engineeringtheir own version, first seen Sep 12, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- bepaldes/heimdall-agent★ 2same name
- RuggG/mcp-engineering★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository8 other files it gives its agents.
- agent-engineeringSkill
- ai-system-designSkill
- break-my-agentSkill
- code-reviewSkill
- llm-evaluationSkill
- python-productionSkill
- rag-engineeringSkill
- security-reviewSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 9
Use this skill for MCP protocol implementations and integrations.
- Distinctive: line 23
Use MCP for exposing tools and context to an AI host.
- Distinctive: line 27
Read protocol and authorization notes for version-sensitive orientation, then verify the target peers against their matching primary …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
MCP Engineering
Use this skill for MCP protocol implementations and integrations. MCP standardizes context and capability exchange; it does not replace application authorization, consent, privacy, or operational controls.
Workflow
- Resolve compatibility first. Inspect the host/client/server, transport, SDK, and protocol revision already in use. Check the matching version of the official specification and SDK migration guidance. Implement only features and extensions supported by both peers. Do not copy examples for another protocol revision or assume session, handshake, transport, or authorization behavior.
- Design a narrow capability surface. Decide whether each capability belongs in a tool (action), resource (context/data), or prompt (reusable template). Keep tools task-specific, bounded, typed, and explicit about side effects. Define input/output schemas, size limits, stable errors, pagination, cancellation, idempotency, and timeouts. Treat tool names, descriptions, annotations, resource content, and model-generated arguments as untrusted; descriptions and annotations are not enforcement controls.
- Enforce identity and authorization at the server. For HTTP transports using MCP authorization, follow the matching protocol authorization requirements and trusted identity-provider configuration; for STDIO and other transports, follow their applicable credential model and security requirements. Validate issuer, audience, expiry, scopes, and resource binding as required by the chosen transport and revision. Authorize every operation against the authenticated principal and tenant at the point of access. Use least-privilege scopes and short-lived credentials. Never forward an incoming token to a downstream service without the required validation and delegation design. Bind opaque task/state handles to the authenticated principal; possession of a handle is not identity.
- Harden inputs and egress. Validate types, lengths, paths, URLs, and resource identifiers on the server. Prevent path traversal, command injection, SQL injection, SSRF, unsafe deserialization, and uncontrolled filesystem/network access. Use parameterized queries, safe process APIs, destination allowlists or controlled egress, and minimal runtime privileges. Require explicit user awareness/authorization for consequential side effects in the host experience.
- Instrument without leaking data. Trace request lifecycle, principal/tenant identifiers that are approved for telemetry, capability name, outcome, latency, rate limits, and dependency failures. Redact tokens, secrets, prompts, retrieved content, and PII by default. Keep audit records protected, access-controlled, and within the approved retention policy.
- Verify protocol and security behavior. Cover matching-version conformance, malformed and oversized payloads, unknown methods/capabilities, authorization failures, cross-tenant access, expired/replayed credentials, malicious resource content, cancellation, timeout, retry/idempotency, and transport failures. Test proxy and OAuth discovery paths for SSRF and token confusion. Load-test using the intended deployment topology. Document supported revisions and deprecations.
- Deploy with bounded access. Pin and track SDK versions, review transitive dependencies, isolate the process, restrict network/filesystem permissions, rotate credentials, and provide health checks, rate controls, alerts, and a rollback plan.
MCP and A2A boundary
Use MCP for exposing tools and context to an AI host. Consider A2A when independent agent systems need a defined task/exchange protocol. Do not conflate the protocols; verify the official current specification and security model for each integration.
Current references
Read protocol and authorization notes for version-sensitive orientation, then verify the target peers against their matching primary specifications and SDK docs. The notes are not a substitute for protocol negotiation or deployment-specific policy.
When Context7 is configured, use it for focused, version-specific SDK documentation after checking the host, server, transport, and protocol versions already in use. Resolve the exact library/version, verify consequential details against the official specification and SDK docs, and keep credentials, secrets, private source code, and restricted data out of queries. Context7 is optional.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.