How do I install the MCP Tool Contracts skill?

Run npx skillmds@latest add mimmo-the-root/mcp-tool-contracts in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is MCP Tool Contracts safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with MCP Tool Contracts?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is MCP Tool Contracts free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published MCP Tool Contracts?

mimmo-the-root (@mimmo-the-root) published this skill. Their other Agent Skills are listed on their SkillMD profile.

MCP Tool Contracts by mimmo-the-root

Name: MCP Tool Contracts
Rating: 4.2 (5 reviews)
Author: mimmo-the-root

MCP Tool Contracts

Canonical, single-source-of-truth mapping of "what I want to do via UEFN's MCP server" to "the exact tool that actually works for it in this kit's setup" — referenced by coder, qa-regression, and project-bootstrap instead of each describing the check in its own words.

File contents

MCP tool contracts

Every agent that talks to UEFN's MCP server needs to do the same handful of things (verify which project is open, list real project content, and so on). This file is the ONE place that says which tool actually does each of those things in this kit's setup — every agent file references this file by name instead of restating the instruction in its own prose.

Why this file exists: a v1.67 incident traced back to the exact same generic instruction ("list a couple of files through the MCP tools") appearing, worded slightly differently, in three separate agent files (coder.md, qa-regression.md, project-bootstrap.md). coder picked AssetTools.find_assets on /Game — the intuitive-sounding choice — which silently returns nothing useful for this kit's custom project content even with the right project open, and reads exactly like "the wrong project is open in UEFN," sending debugging in the wrong direction. Fixing the wording in all three files at once worked, but it relied on remembering to touch all three every time this file's contents change. A single referenced file can't diverge from itself.

Contract: verifying which project UEFN has open

Use: ValkyrieToolset.VerseToolset.ListFiles

Don't use: AssetTools.find_assets on /Game — it silently returns nothing useful for this kit's custom project content even when the correct project is open in UEFN. This is a configuration-specific quirk of this MCP/UEFN setup, not documented anywhere in Epic's own docs as of this writing — see the entry in ~/.claude/skills/uefn-lessons/SKILL.md's "MCP / tooling quirks" for the original discovery.

Procedure: list the project's real Verse/asset content with ValkyrieToolset.VerseToolset.ListFiles and check the path/name against the value already written at the top of the project's CLAUDE.md, in the "Project identity" section — never the current folder's name (Content, identical across every project). If the listing doesn't match, or is ambiguous, STOP and warn the owner: UEFN probably has a different project open. Every agent that touches MCP before writing/modifying anything runs this check first (coder, qa-regression before a play-session, project-bootstrap during initial analysis).

Adding a new contract

When a new MCP-tool gotcha gets discovered (a tool that seems right by name but doesn't do what's expected, or two tools that both technically work but one is clearly the intended one for this kit), add it here as a new ## Contract: section with the same three fields — Use / Don't use / Procedure — rather than describing it inline in whichever agent file happened to discover it. Also add a short entry to ~/.claude/skills/uefn-lessons/SKILL.md if the gotcha would help on a different island, not just this project (see that skill's own guidance on the "would this help on a different island?" test).

---name: mcp-tool-contractsdescription: Canonical, single-source-of-truth mapping of "what I want to do via UEFN's MCP server" to "the exact tool that actually works for it in this kit's setup" — referenced by coder, qa-regression, and project-bootstrap instead of each describing the check in its own words.---# MCP tool contractsEvery agent that talks to UEFN's MCP server needs to do the same handful of things (verify whichproject is open, list real project content, and so on). This file is the ONE place that says whichtool actually does each of those things in this kit's setup — every agent file references thisfile by name instead of restating the instruction in its own prose.**Why this file exists**: a v1.67 incident traced back to the exact same generic instruction("list a couple of files through the MCP tools") appearing, worded slightly differently, in threeseparate agent files (`coder.md`, `qa-regression.md`, `project-bootstrap.md`). `coder` picked`AssetTools.find_assets` on `/Game` — the intuitive-sounding choice — which silently returnsnothing useful for this kit's custom project content even with the right project open, and readsexactly like "the wrong project is open in UEFN," sending debugging in the wrong direction. Fixingthe wording in all three files at once worked, but it relied on remembering to touch all threeevery time this file's contents change. A single referenced file can't diverge from itself.## Contract: verifying which project UEFN has open**Use:** `ValkyrieToolset.VerseToolset.ListFiles`**Don't use:** `AssetTools.find_assets` on `/Game` — it silently returns nothing useful for thiskit's custom project content even when the correct project is open in UEFN. This is aconfiguration-specific quirk of this MCP/UEFN setup, not documented anywhere in Epic's own docs asof this writing — see the entry in `~/.claude/skills/uefn-lessons/SKILL.md`'s "MCP / toolingquirks" for the original discovery.**Procedure:** list the project's real Verse/asset content with `ValkyrieToolset.VerseToolset.ListFiles`and check the path/name against the value already written at the top of the project's `CLAUDE.md`,in the "Project identity" section — never the current folder's name (`Content`, identical acrossevery project). If the listing doesn't match, or is ambiguous, STOP and warn the owner: UEFNprobably has a different project open. Every agent that touches MCP before writing/modifyinganything runs this check first (`coder`, `qa-regression` before a play-session, `project-bootstrap`during initial analysis).## Adding a new contractWhen a new MCP-tool gotcha gets discovered (a tool that seems right by name but doesn't do what'sexpected, or two tools that both technically work but one is clearly the intended one for thiskit), add it here as a new `## Contract:` section with the same three fields — Use / Don't use /Procedure — rather than describing it inline in whichever agent file happened to discover it. Alsoadd a short entry to `~/.claude/skills/uefn-lessons/SKILL.md` if the gotcha would help on adifferent island, not just this project (see that skill's own guidance on the "would this help ona different island?" test).

Frequently asked questions