skills.rest
leroystolberg-hash
leroystolberg-hashCOMMUNITY
Explore all skills by this author →
GitHub

TVL

Translates virtual actions into concrete MCP tool calls via a caching and resolution pipeline.

Updated Mar 10, 2026
Search Tags:#typescript#caching#mcp#embedding search#auth routing#tool virtualization#chain compilation
One-click install
npx skills add https://github.com/leroystolberg-hash/TVL --skill tvl
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: TVL
Source: https://github.com/leroystolberg-hash/TVL/tree/main
Command: npx skills add https://github.com/leroystolberg-hash/TVL --skill tvl

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps you implement and operate the Tool Virtualization Layer (TVL), which converts many MCP tool schemas into 7 stable virtual actions so an LLM can call tools without context bloat.

Core Features & Use Cases

  • 7 virtual actions interface: send, fetch, store, schedule, search, transform, monitor to keep LLM tool usage stable across servers.
  • TLB + MMU resolution pipeline: fast repeated-call translation with a cache (TLB) and deterministic/LLM-assisted resolution (MMU).
  • Tool registry + embeddings: register MCP servers, index tool schemas, and match capabilities using embedding search with schema validation.
  • Auth-aware routing: credential scoring and OAuth/API-key flows, ensuring only authorized tools are selected.
  • Chain compilation & execution: detect repeated call sequences and compile them into atomic chains for efficiency.

Quick Start

Give the system the goal “Send an email to [email protected] with the message ‘Hello’ using TVL with one local MCP server,” then ask it to register the server if needed and execute the corresponding virtual action end-to-end.

Frequently Asked Questions about TVL

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I stop MCP tool schemas from bloating LLM context when managing many servers?▼

Prevent MCP schema context bloat by virtualizing tool calls into 7 stable actions. The layer translates virtual actions into concrete MCP calls internally, keeping the LLM context stable across servers.

What is the best way to map LLM tool calls to MCP servers without exposing schemas?▼

Virtualize tool calls into 7 stable actions using a 4-phase MMU pipeline with entity resolution and schema validation. This translates LLM requests into concrete MCP calls without exposing schemas.

How do I implement auth-aware routing for MCP tool selection?▼

Implement auth-aware routing for MCP tool selection using credential scoring alongside OAuth/API-key flows. This ensures the virtualization layer selects only authorized tools during capability matching.

Can I cache repeated MCP tool calls in TypeScript for faster execution?▼

Cache repeated MCP tool calls using a TLB cache with persistence. This provides fast repeated-call translation within the TypeScript strict ESM-compatible runtime environment.

Does this tool virtualization approach support compiling repeated MCP call sequences?▼

Yes, the tool virtualization approach supports chain compilation by detecting repeated MCP call sequences and compiling them into atomic chains, resulting in more efficient execution.

Do I need TypeScript and Vitest to use the tool virtualization layer?▼

Yes, you need a TypeScript strict ESM-compatible environment with Vitest coverage thresholds. The implementation requires the official MCP SDK for transport and Vitest for testing.

Related Skills

Explore More Software Engineering Skills→
thistleknotthistleknot

mcp-tool-registry

Discover, register, and route MCP tools across multiple servers.

Community
Intermediate
gtriggtrig

mcp-go-sdk

Build MCP servers and tools using the official Go SDK.

Community
Advanced
mbadozmbadoz
2

regex-vs-llm-structured-text

Decide between regex and LLM parsing for structured text extraction.

Community
Advanced
JDL440JDL440

Agent Tool-Calling Architecture Pattern

Define a centralized tool registry for agent tool-calling with per-agent allowlists.

Community
Advanced