mcp-server-spec

mcp-server-spec is a cursor rule for Cursor from mohitagw15856/pm-claude-skills. It costs 96 tokens per session (1,239 once invoked), scanned A, original, MIT.

A design plan for an MCP server, which exposes a product's capabilities as tools that AI agents can use. It defines the tools, inputs, results, errors, authentication, permissions, and safety limits.

In plain words
What is it for?
It helps turn user tasks into a small toolset, define what each tool can do, restrict credentials, list excluded operations, and test agent use.
Why use it?
An API built for human developers can overwhelm an agent with too many low-level choices or expose risky actions without clear boundaries.

Cursor rule for Cursor

Written for Cursor: a Cursor rule (.mdc).

Good fit It helps turn user tasks into a small toolset, define what each tool can do, restrict credentials, list excluded operations, and test agent use.

Compare 6 cursor rules from other repositories ↓
Install with agentmods
npx agentmods add rules/mohitagw15856/pm-claude-skills/mcp-server-spec
About the project

PM Skills is a collection of plain-Markdown instructions that teach AI assistants structured methods for handling professional, personal, and life-admin tasks. People use it with Claude, ChatGPT, Gemini, Cursor, Codex, and other supported agents for work such as writing product requirements, reviewing documents, or planning difficult situations.

mohitagw15856/pm-claude-skills · 1,352 stars · on GitHub · mohitagw15856.github.io

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.

Clone the repo
git clone --depth 1 https://github.com/mohitagw15856/pm-claude-skills

Made for: Cursor.

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-server-spec

README.md
[![agentmods](https://agentmods.dev/badge/rules/mohitagw15856/pm-claude-skills/mcp-server-spec/github.svg)](https://agentmods.dev/rules/mohitagw15856/pm-claude-skills/mcp-server-spec)
Your own site
<a href="https://agentmods.dev/rules/mohitagw15856/pm-claude-skills/mcp-server-spec"><img src="https://agentmods.dev/badge/rules/mohitagw15856/pm-claude-skills/mcp-server-spec/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-server-spec

Your own site · 80×15
<a href="https://agentmods.dev/rules/mohitagw15856/pm-claude-skills/mcp-server-spec"><img src="https://agentmods.dev/badge/rules/mohitagw15856/pm-claude-skills/mcp-server-spec.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 96 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,239 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 $0.00096 $0.01239
Opus 5 $0.00048 $0.00620
Sonnet 5 $0.00019 $0.00248
Haiku 4.5 $0.00010 $0.00124

Measured 7d ago against content hash 2f6bdbde51f5, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

mcp-server-spec 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 7d 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.

exports/cursor/pm-agentnative/mcp-server-spec/mcp-server-spec.mdc · 72 lines

How it starts

The opening of the file, as written. The whole thing — 72 lines — stays where its author put it; the contents beside it link to each section on GitHub.

MCP Server Spec Skill

Every SaaS is shipping an MCP server; most dump their REST API as forty tools and wonder why agents flail. This skill designs the server as what it actually is: a user interface for a non-human user — few tools, task-shaped, with descriptions written for a model deciding under uncertainty.

What This Skill Produces

  • A toolset design: 3-10 tools mapped to agent tasks, not API endpoints
  • Per-tool specs: name, description (the routing surface), parameters, returns, error behaviour
  • Auth & scoping decisions: how credentials flow, what a token can never do
  • An explicit not-exposed list with reasons — the most load-bearing section
  • A test plan: the agent-eval loop that proves the toolset works

Required Inputs

Ask for (if not already provided):

  • The product and what users hire it for (the top 5 jobs, not the feature list)
  • The existing API surface (endpoints or capability list) if one exists
  • Who the agent acts for — the end user's own account? a service account? multi-tenant?
  • The riskiest actions the product supports (deletes, sends, payments, permission changes)

Design Method

  1. Start from agent tasks, not endpoints. List the 5-8 things an agent will actually be asked to do with this product ("file an expense", "find last quarter's report", "summarise ticket history"). Each becomes one tool — even if it spans four API calls internally. An endpoint-mirrored toolset makes the agent do your orchestration; a task-shaped one does it for them.
  2. Keep the toolset small. Every tool dilutes selection accuracy on every call. Target ≤10; past ~15, split into separately-loadable servers by workflow. Merge list/get/search variants behind one tool with parameters where natural.
  3. Write descriptions as routing surfaces. The description is all the model sees when choosing. Formula per tool: what it does (one clause) · when to use it and when to use the sibling tool instead · what it returns. Test: could a model pick correctly between your two closest tools from descriptions alone?
  4. Design returns for context windows. Return the 6 fields an agent needs, not the 60 the API has; include stable IDs for chaining; paginate with explicit has_more; keep any response under ~2k tokens by default with an opt-in for detail.
  5. Make errors instructive. An agent retries what it understands: "date must be YYYY-MM-DD" beats 400 Bad Request. Every error names the parameter at fault and the fix.
  6. Draw the safety boundary. Classify every capability: expose (read/create, low blast radius) · expose gated (destructive/outward-facing — require an explicit confirmation parameter and document that clients should surface approval) · never expose (auth changes, deletes without recovery, bulk exports of other users' data). The never-list ships in the spec with reasons.
  7. Specify auth honestly. OAuth per end user (agent acts as the user, inherits their permissions) vs API key (service account — then per-tool scoping matters more). State token lifetime, revocation, and what happens mid-session on expiry.

Read the full file on GitHub · 72 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. 7d ago First seen · 72 lines · 96 tokens per session scan A 2f6bdbde51f5

Subscribe to this mod's changes

mcp-server-spec is a cursor rule published in the GitHub repository mohitagw15856/pm-claude-skills (1,352 stars, last pushed 2d ago), licensed MIT. It adds 96 tokens to every session and 1,239 once invoked, about $0.0005 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 2026-09-03.