Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Local Llm Troubleshooter

    2

    Diagnose and fix broken local LLM stacks, GPU issues, and stalled model downloads across Ollama, LM Studio, and more.

    Free

    9 usesSecurity scanned

    local-llm-troubleshooter

    Example session with this skill installed

    My Ollama server is responding but I can't get Llama 3 70b to load, it just hangs. Can you run the doctor script and tell me why it's stuck?

    • Read your context and instructions
    • Compiled the local-llm-troubleshooter

    Verdict: STUCK (Ollama)
    Cause: CUDA Out of Memory (OOM) detected in logs.
    Model: Llama-3-70b (Q4_K_M)

    Fix

    1. Set OLLAMA_NUM_PARALLEL=1
    2. Reduce context window from 8k to 4k in Modelfile.
    3. If issue persists, switch to Q2 quantization.
      Verification: Re-run diagnostic to confirm GPU runner load.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    What it does

    The Local LLM Troubleshooter is a diagnostic power-tool for developers and AI engineers whose local inference stacks (Ollama, LM Studio, llama.cpp, vLLM, or Hugging Face) are failing. It eliminates the guesswork of "why is my model slow?" or "why won't this load?" by running a bundled diagnostic script that probes connection states, scans logs for failure signatures, and detects stalled downloads.

    Why use this skill

    Prompting a generic AI about local hardware issues often leads to circular advice. This skill is better because it uses llm_doctor.py to act as a system sensor. It identifies specific technical blockers like GGUF version mismatches, CUDA OOM (Out of Memory) errors, port conflicts, runner crashes, and stalled Hugging Face blobs. It maps these findings to a curated playbook of OS-specific fixes for Apple Silicon, NVIDIA, and WSL2 environments.

    Supported tools

    • Inference Servers: Ollama, LM Studio, llama.cpp, vLLM
    • Model Sources: Hugging Face (hub downloads), Ollama library
    • Frameworks: GGUF, local runners, GPU-accelerated backends

    What the output looks like

    The skill provides a structured triage report including a connectivity verdict (up/down/stuck), identification of the specific bottleneck, and an ordered list of high-probability fixes—ranging from context window adjustments to environment variable corrections.

    • Cannot auto-apply fixes; user must run commands manually.
    • Requires Python 3 to run the diagnostic script.
    • Limited visibility into proprietary, closed-source inference engines.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    9 uses

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 5 months ago

    • Free with an Agensi account

    Requires

    Python

    Permissions

    Environment Variables
    Listed5 months ago

    What's inside

    Frequently Asked Questions

    This skill is purpose-built to diagnose common local inference failures, such as CUDA OOM (Out of Memory) errors, broken GGUF model paths, stalled Hugging Face downloads, and port conflicts across stacks like Ollama and LM Studio. It replaces manual log-scouring with a structured diagnostic report and OS-specific fixes.

    The skill currently supports macOS (Apple Silicon), Linux (NVIDIA/AMD), and Windows via WSL2. It is designed to work with any AI agent that has file system and terminal execution permissions to run the included diagnostic sensor.

    No complex installation is required; the package includes the `llm_doctor.py` diagnostic script and the agent logic. Once imported into your agent's environment, it is ready to scan your local setup immediately.

    Your purchase includes the diagnostic logic, the `llm_doctor.py` system sensor, and access to a curated playbook of fixes for the major inference engines like vLLM, llama.cpp, and Ollama.

    Yes, users receive lifetime access to version updates, ensuring the troubleshooter stays current with new releases of popular inference tools and changes in Hugging Face download protocols.

    New in Agents & Orchestration