agentleFS
HumanleapHumanleap/agent-skills/skills/toolrouter/SKILL.md

toolrouter

Skill0 starsChanged 41 days agoMITView on GitHub

Discover and run ToolRouter tools through one MCP connection for web search, scraping, SEO research, image and video generation, audio, company and economic data, and supported connected-account actions including X and LinkedIn posting. Use when ToolRouter is connected or the user chooses it for an external task.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Discover first: Use returned operation schemas.

    line 17
  2. Very specific:

    Discover the outcome: Call discover with a query describing the job: research a company, check keyword demand, extract a website, generate …

    line 28
  3. Very specific:

    Execute: Call use_tool with that tool, skill and schema-matching input.

    line 30
  4. Very specific:

    Resolve asynchronous work: Follow the returned job handle with the advertised status tool, such as job_get, until outputs are ready or a …

    line 31

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

ToolRouter

One connection gives an assistant access to hundreds of hosted tools and provider operations. Discover the right operation before calling it; the live catalog determines availability, input fields and account requirements.

Installation

Install this skill with pnpm dlx skills add Humanleap/agent-skills --skill toolrouter. Connect the client's remote MCP server to https://api.toolrouter.com/mcp using its authentication flow. Setup: https://toolrouter.com/docs/markdown/quickstart.

Four Hard Rules — Read First

  1. Discover first. Use returned operation schemas; never invent a tool, provider connection or input field.
  2. Respect the chosen provider. A capability request does not authorise connecting another account, changing billing, or replacing a service the user specified.
  3. Check action and spending authority. Posting, deleting, messaging and paid operations require authority for that task. A catalog entry does not establish access to the user's accounts.
  4. Finish the actual job. A queued job is unfinished. Check status before repeating an uncertain write and report provider errors or missing credentials plainly.

Authentication

Read tools/list for the current client profile. Follow the advertised OAuth or account-claim flow and keep credentials in the client credential store. Do not include tokens in chat, skills or shared configuration. Some operations use hosted providers; others require a connected provider account. Check the operation before promising access.

Core Workflow

  1. Discover the outcome. Call discover with a query describing the job: research a company, check keyword demand, extract a website, generate an image, transcribe media, or publish an authorised social post.
  2. Read the operation. Select the returned tool and skill; inspect the exact input schema, price, account requirements and limitations.
  3. Execute. Call use_tool with that tool, skill and schema-matching input. Built-in MCP tools are called directly. Check credits_balance when the task needs paid credits.
  4. Resolve asynchronous work. Follow the returned job handle with the advertised status tool, such as job_get, until outputs are ready or a blocker is reported.
  5. Return the useful result. Share source links, completed assets or the provider's verified action result, with relevant dates and limitations.

Essential Tools

ToolUse
discoverFind operations, schemas and examples for the requested outcome or exact tool.
use_toolExecute one discovered provider operation.
credits_balanceInspect the account's credit balance when available.
job_getCheck an asynchronous job when the returned response supplies a handle for it.

Inspect tools/list before relying on a built-in name: client profiles and the catalog can change.

Common Patterns

  • Research and extraction: discover web search or scraping, use original source URLs, then extract the fields the user needs.
  • SEO: discover keyword, SERP, backlink or competitor operations; retain provider, market, time range and measurement limits.
  • Creative work: discover the model or media operation, respect the approved budget, then poll and return completed output URLs.
  • Business and economic data: select the original registry or statistical source and retain dates, units and geography.
  • Social publishing: discover the exact account operation. Supported catalog examples include linkedin-post.post_text, linkedin-post.post_link, linkedin-post.post_image, x-manager.post_tweet and x-manager.post_image_tweet. Confirm the connected account and authorised content from the request; discover the current schema before executing.

Providers and Supporting Resources

Read the provider reference when a task names a provider. It lists all 55 entries in the public provider index as of 3 October 2026, including Firecrawl, Exa, Serper, DataForSEO, OpenAI, Anthropic, fal.ai, ElevenLabs, Google, OpenRouter, Companies House, SEC EDGAR and the World Bank. Public provider pages describe integrations; they do not prove every operation is deployed or that the user has access.

Common Gotchas

  • The website's provider count and the deployed tool count can differ. Use the live operation schema.
  • A social connection does not imply support for every social platform or permission to post.
  • Some client-specific MCP profiles expose fewer operations. Do not promise media generation in a profile that filters it out.
  • An empty search result is a coverage limit, not proof of absence.

Quick Reference

Connect → discover → inspect schema and authority → use_tool → resolve job → share verified result.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.