agentsclimarketplace

Tool agnostic mcp routing

Skill ntaffzii/Skill-Agents/Skill.md/tool-agnostic-mcp-routing

Route tasks to available MCP or agent tools by capability rather than exact local tool names. Use when an agent has tools from Gemini, ChatGPT, Claude, Codex, or another MCP server and must decide which available tool capability to use.From its SKILL.md

Install
npx -y skills add ntaffzii/Skill-Agents --skill tool-agnostic-mcp-routing

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 4 stars4 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

6.3 KB, ~1.2k tokens by cl100k_base, as published. Nobody here has run it

Tool-Agnostic MCP Routing

Use this skill when the current environment may have different tool names than this repository.

Capability Map

Translate task needs into capabilities:

NeedCapability
Inspect codefile search, file read
Choose tool scopetoolset recommender, toolset inspector, registry
Edit codepatch, file write, code edit
Verify codecommand runner, test runner
Hand off blocked commandsuser-run script generator, command handoff planner, script risk validator
Plan validationvalidation planner, command policy checker
Inspect repository stateversion control status, diff, log
Control repository statebranch creator, checkout, stage, unstage, commit
Understand project shapeproject stack, scripts, manifest inspection
Understand project docsdocumentation index, README/docs reader, context bundle
Build repo maprepository index builder, symbol/file search, related-file finder
Compare external MCP ideasMCP catalog search, repo pattern comparison, local adaptation planner
Inspect dependenciespackage manager detection, dependency manifest reader, lockfile status
Inspect containersDocker CLI checker, Dockerfile inspector, Compose service inspector, Docker validation planner
Inspect API surfaceroute finder, endpoint extractor, OpenAPI spec reader, API config hints
Inspect data layerschema finder, migration finder, ORM model extractor, database config hints
Query configured databaseread-only SQL planner, SQL risk explainer, read-only Postgres runner
Inspect configurationconfig finder, env key lister, secret hygiene checker
Inspect teststest file finder, framework detector, source-to-test mapper
Inspect CICI file finder, GitHub Actions job inspector, validation command extractor
Inspect local tasksTODO/FIXME scanner, roadmap file finder
Inspect dependency riskunpinned dependency finder, high-risk dependency name checker
Scan repository securitysecret scanner, dangerous command scanner, env exposure scanner, security report
Inspect release readinesschangelog finder, version detector, release checklist
Snapshot before broad editsbackup planner, zip snapshot creator
Edit structured registriesJSON/YAML/TOML reader, path getter, structured path patcher
Preserve project memorymemory saver, memory search, project memory summary
Build personal contexttyped memory saver, context pack builder, memory handoff generator
Search long-running contextvector or semantic memory search, related-memory finder, memory cluster summarizer
Build RAG contexttext chunker, RAG provider planner, embedding request planner, source metadata preserver
Inspect GitHub metadataworkflow finder, Actions inspector, PR description drafter
Inspect GitHub API datarepo reader, issue reader, PR reader, PR file lister, check status reader
Inspect browser/UI readinessbrowser capability checker, static HTML inspector, smoke-check planner
Map a web pageHTML page mapper, heading/link/form/button extractor, element finder by label
Inspect live UIPlaywright runtime checker, page inspector, screenshot capture
Automate live UIPlaywright click, fill, text assertion, console/network failure inspector, accessibility snapshot, persistent session runner
Inspect design filesFigma auth checker, design file summary, token/component extractor, frontend implementation planner
Try safe experimentssandbox workspace creator, snippet compiler, sandbox run planner
Understand tool failuresaudit log, policy denial inspection
Audit MCP safetyrisk classifier, mutating tool finder, policy coverage checker
Research current factsweb search, URL fetch
Capture public web pagesprovider-neutral static fetch, browser capture, optional cloud scraper adapter, social public-only reader
Research finance marketsmarket quote reader, crypto price reader, finance-news planner, watchlist summarizer, position-risk calculator
Work with NotionNotion auth checker, page search, page reader, note/update payload planner
Bridge Obsidian and NotionObsidian note inspector, Notion payload planner, Markdown conversion planner, sync checklist
Work with chat appsSlack/Discord config checker, message search, channel summarizer, reply drafter, action-item extractor
Triage inboxemail config checker, email search planner, email summarizer, action extractor, reply drafter
Plan calendarcalendar config checker, event summarizer, daily plan builder, meeting prep drafter, follow-up extractor
Plan issue workissue reference parser, issue draft builder, task breakdown, issue update planner
Inspect UIbrowser automation, screenshot
Work with images/videomedia inspection, frame extraction
Improve promptsprompt analysis, rewrite, scoring
Save progresshandoff, document write

Workflow

  1. Identify the next job

    • What must be learned, changed, verified, or created?
  2. Choose capability

    • Pick the needed capability before choosing a tool.
    • If several tools can do it, prefer the least risky one.
  3. Map to current tools

    • Read available tool names and descriptions.
    • If registry/introspection tools exist, call them before guessing.
    • Select the closest matching tool.
    • If no tool exists, explain the missing capability and use a text-only fallback.
  4. Execute safely

    • Inspect before editing.
    • Preview before broad writes when possible.
    • Run narrow validation before broad validation.
  5. Record mapping

    • In the final answer, say what capability was used.
    • Mention exact tool names only when useful to the user.

Rules

  • Do not assume this repo's mcp-tools/ exists.
  • Do not call tools by guessed names.
  • Do not use a high-risk tool when a read-only tool is enough.
  • If tool descriptions are ambiguous, choose the safer read-only action first.

Output Format

End with:

  • Needed capability
  • Tool used or fallback
  • Result
  • Verification

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 326,790. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.