mcp-architect-lord

mcp-architect-lord is a skill for Claude Code, Codex from m3taz-ahmed/ai-globals. It costs 49 tokens per session (1,516 once invoked), scanned A, original, MIT.

A guide for designing and securing MCP servers. MCP is a protocol that lets AI agents use tools, data, and APIs; the guide covers stateless server design, OAuth 2.1 access control, protection against malicious tool descriptions, and protocol migration.

In plain words
What is it for?
It is for planning or migrating MCP servers, including tool naming, external session storage, OAuth resource-server behavior, and compatibility between protocol versions.
Why use it?
It addresses risks such as leaked secrets, unsafe tool definitions, failed authentication, and servers that cannot scale or recover cleanly. It also helps handle breaking changes in the specified MCP version.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one.

Good fit It is for planning or migrating MCP servers, including tool naming, external session storage, OAuth resource-server behavior, and compatibility between protocol versions.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/m3taz-ahmed/ai-globals/mcp-architect-lord
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add m3taz-ahmed/ai-globals --skill mcp-architect-lord
Clone the repo
git clone --depth 1 https://github.com/m3taz-ahmed/ai-globals

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for mcp-architect-lord

README.md
[![agentmods](https://agentmods.dev/badge/skills/m3taz-ahmed/ai-globals/mcp-architect-lord/github.svg)](https://agentmods.dev/skills/m3taz-ahmed/ai-globals/mcp-architect-lord)
Your own site
<a href="https://agentmods.dev/skills/m3taz-ahmed/ai-globals/mcp-architect-lord"><img src="https://agentmods.dev/badge/skills/m3taz-ahmed/ai-globals/mcp-architect-lord/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for mcp-architect-lord

Your own site · 80×15
<a href="https://agentmods.dev/skills/m3taz-ahmed/ai-globals/mcp-architect-lord"><img src="https://agentmods.dev/badge/skills/m3taz-ahmed/ai-globals/mcp-architect-lord.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 49 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,516 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00049 $0.01516
Opus 5 $0.00024 $0.00758
Sonnet 5 $0.00010 $0.00303
Haiku 4.5 $0.00005 $0.00152

Measured 6d ago against content hash 9ba541f74d21, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-12, from the pricing page.

Security

Grade A, and why

mcp-architect-lord scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/mcp-architect-lord/SKILL.md · 64 lines

How it starts

The opening of the file, as written. The whole thing — 64 lines — stays where its author put it; the contents beside it link to each section on GitHub.

MCP Architect Lord

[OBJ] Design, secure, and migrate MCP servers compliant with the 2026-07-28 protocol specification.

Problem

MCP servers are the bridge between LLM agents and real-world tools, data, and APIs. A poorly designed server leaks secrets, accepts poisoned tool definitions, breaks under transport changes, or fails OAuth handshake — turning the integration surface into an attack surface. The 2026-07-28 spec introduces breaking changes (CIMD, Resource Server, FastMCP v4) that require deliberate migration, not patch-on-patch.

Rules

  1. [REQ] Stateless core. The server core (tool dispatch, resource reads, prompt rendering) MUST be stateless. Session state lives in external stores (Redis, DB) keyed by session ID, never in process memory. Enables horizontal scaling and crash recovery.
  2. [REQ] Tool naming conventions. Tool names MUST be snake_case, verb_noun format, ≤32 characters. Examples: get_issue, create_ticket, search_docs. No camelCase, no kebab-case, no abbreviations beyond well-known ones (id, url, api).
  3. [REQ] OAuth 2.1 Resource Server. The MCP server acts as a Resource Server per RFC 9728. Protected Resource Metadata is served at /.well-known/oauth-protected-resource. Authorization Server metadata discovered via issuer field. No hardcoded auth server URLs.
  4. [REQ] CIMD not DCR. Use Client-Initiated Metadata Discovery (CIMD) per RFC 9728. Do NOT use Dynamic Client Registration (DCR) — deprecated in 2026-07-28. Clients discover metadata from the protected resource, not by registering.
  5. [REQ] PKCE on all flows. Proof Key for Code Exchange (RFC 7636) MUST be enforced on every authorization code flow, including confidential clients. No response_type=code without code_challenge.
  6. [REQ] DPoP for sender-constrained tokens. Demonstrating Proof-of-Possession (RFC 9449) binds access tokens to the client's key. Prefer DPoP over bearer tokens for high-security tools. Resource Server MUST validate DPoP proof on each request.
  7. [REQ] Dual-era support. Servers MUST support both legacy (2025-11-25) and current (2026-07-28) protocol versions simultaneously. Detect era from protocolVersion in _meta. Modern path = stateless. Legacy path = backward-compat shim. Deprecation timeline: 12 months from spec release.
  8. [REQ] Tool poisoning defense. Hash-pin tool definitions at registration time. On each load, verify the hash matches. A mismatch (definition tampering) = BLOCK + alert. Never execute a tool whose definition hash differs from the pinned version.
  9. [REQ] Per-tool scopes. Each tool declares required OAuth scopes in its definition (annotations.scopes). The Resource Server validates the access token has ALL required scopes before dispatch. No blanket scope access.
  10. [REQ] MRTR pattern. Model-Reads-Tool-Returns: tools return structured data (JSON), not prose. The model interprets results; the tool does not narrate. Enables caching, validation, and replay.
  11. [REQ] Extensions framework. The server MUST support the extensions framework: tasks (long-running async), apps (UI surfaces), skills (declarative capabilities), EMA (Enterprise Management API). Extensions are opt-in via capability negotiation.
  12. [REQ] Cacheable listings. tools/list, resources/list, prompts/list responses MUST include ETag and Last-Modified headers. Clients use conditional requests (If-None-Match). Reduces bandwidth and latency on large catalogs.
  13. [REQ] Header-based routing. Use Mcp-Method and Mcp-Name headers for routing and rate-limiting, not URL paths or query params. Enables clean reverse-proxy and load-balancer configurations without parsing JSON bodies. Mcp-Session-Id is REMOVED in 2026-07-28 — do not use it.
  14. [REQ] One server = one domain = one auth boundary. A single MCP server serves one logical domain (e.g., "GitHub issues" or "Linear tickets"). Do not mix domains in one server — it breaks scope granularity and audit clarity.
  15. [REQ] CVE tracking. Monitor CVEs for the MCP SDK, transport libraries, and OAuth libraries in use. Subscribe to security advisories. Patch within 72h for critical, 7d for high.
  16. [REQ] Inspector testing. Every server MUST pass the MCP Inspector test suite before deployment. Run mcp-inspector locally during development; run it in CI before merge. No server ships without a green Inspector run.
  17. [REQ] Client config generation. The server MUST auto-generate client configuration snippets (Claude Desktop, VS Code, Cursor) from its own metadata. No hand-written JSON configs that drift from the server's actual capabilities.
  18. [REQ] Registry registration. Publish server to an MCP registry (local or remote) with: name, version, description, transport, auth metadata, tool count, health endpoint. Enables discovery and governance.
  19. [REQ] Transport selection. Use stdio for local single-user tools (CLI integrations, local file access) — stdio does NOT follow OAuth auth spec. Use Streamable HTTP (single POST endpoint, per-request SSE stream) for remote multi-user services. HTTP+SSE transport is DEPRECATED — do not use for new servers. Document the rationale. Never expose stdio servers over network without a wrapper.
  20. [PROHIBIT] Shipping an MCP server without OAuth Resource Server metadata, tool hash pinning, and Inspector validation — these three are non-negotiable for production.

Read the full file on GitHub · 64 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 6d ago First seen · 64 lines · 49 tokens per session scan A 9ba541f74d21

Subscribe to this mod's changes

mcp-architect-lord is a skill published in the GitHub repository m3taz-ahmed/ai-globals (5 stars, last pushed yesterday), licensed MIT. It adds 49 tokens to every session and 1,516 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-06.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

local-ai-agents

Build local-first AI agents that run entirely on a developer workstation with Microsoft Foundry Local and Qwen function-calling models. Covers Small Language Models (SLMs), the OpenAI-compatible local endpoint, sandboxed local tools, local RAG with Chroma, local MCP servers, hybrid cloud/local routing, and the…

microsoft/ai-agents-for-beginners · 200 tokens

next-cache-components-adoption

Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…

vercel/next.js · 95 tokens

chat-pet-sprite-creation

Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.

microsoft/vscode · 53 tokens

cpu-profile-analysis

Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…

microsoft/vscode · 71 tokens

insight-error-page

Write or audit an insight-kind error page for the Next.js dev overlay. Use when creating a new errors/ .mdx page, auditing an existing one, or checking that a page matches the framework fix cards. Covers page structure, title alignment, FixCard cards with Copy prompt button, code snippets, terminology verification…

vercel/next.js · 83 tokens