Agent Stall Watchdog for Claude Code — Catch Silent Stalls

    1

    Monitor file evidence from AI agents and catch stalls that status labels hide.

    $29

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Agent Stall Watchdog for Claude Code — Catch Silent Stalls

    Example session with this skill installed

    I have an agent working on a large dataset. Check its progress evidence against the last snapshot to see if it has stalled.

    • Read your context and instructions
    • Compiled the agent stall watchdog

    The watchdog has completed the check across 3 monitored agents. Two agents are healthy and making observable progress, while one agent has exceeded its stall window despite a 'working' status label.

    Agent Stall Report

    Agent IDDeclared StateLast ChangeStatusEvidence Check
    data-scrapper-01working2m ago✅ HEALTHYHash & Size changed
    indexer-mainworking45m ago❌ STALLEDNo file changes for 30m
    log-rotatoridle5m ago✅ HEALTHYWithin stall window

    Analysis

    The agent indexer-main is currently flagged. Although the system process is active and the state label is set to working, the watched artifact output/indices.json has not changed in size or content hash for 45 minutes, exceeding the 30-minute threshold. This suggests the agent may be stuck in a network timeout or a logic loop.

    Next steps

    1. Inspect the logs for indexer-main to identify the block point.
    2. Restart the stalled agent process and clear its local cache.
    3. Run a new snapshot command once the agent resumes normal operation.

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

    About this skill

    Agent Stall Watchdog is a small file-evidence monitor for Claude Code, terminal agents, and long-running automation loops.

    It compares watched report and artifact files across snapshots using modification time, size, and content hash. If an agent still says it is working but none of its evidence files have changed within the stall window, the watchdog reports it as stalled.

    What you get:

    • Snapshot capture for report and artifact files
    • Stall detection based on file evidence, not state labels
    • Missing-file and unwatched-agent detection
    • Setup error classification when zero agents are loaded
    • Runnable demo with one fresh agent and one stalled agent

    Ideal for developers supervising multi-agent pipelines, batch jobs, recurring automations, and AI workflows where "in progress" is not proof of progress.

    Requires direct file system access to agent artifacts. Cannot monitor standard output unless redirected to a file. Requires manual configuration of stall windows.

    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

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    Requires

    Python

    Permissions

    Terminal / Shell
    Listed1 month ago

    What's inside

    Frequently Asked Questions

    Traditional monitors rely on an agent's own status reports, which can be inaccurate if the agent is stuck in a loop. This skill uses deterministic Python logic to verify physical file changes and content hashes, providing a 'ground-truth' check independent of the agent's internal reasoning.

    The watchdog is compatible with any AI agent or autonomous framework that produces file-based artifacts, logs, or reports. As long as the agent writes to a file system accessible by the watchdog script, it can be monitored regardless of the underlying LLM or framework.

    You receive the core Python monitoring engine, a sample `watchdog_config.json` for easy setup, and documentation detailing how to integrate the exit codes into your existing CI/CD or alerting pipelines.

    Setup involves defining your target file paths and stall thresholds (e.g., 5 minutes) in the configuration file. Once configured, you simply run the script to establish a baseline snapshot and schedule regular checks via your preferred task runner.

    No, the watchdog is designed to be lightweight. It performs quick file system metadata checks and hash comparisons, ensuring it doesn't consume significant CPU or I/O resources that your primary agents need to operate.

    New in Agents & Orchestration