Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Autonomous Supervisor

    1

    Autonomous Supervisor gives an AI coding agent the working habits of a senior lead.

    $10

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Autonomous Supervisor

    Autonomous Supervisor

    Example session with this skill installed

    I need to add a CSV export feature to this tool. It should handle piped input and escape special characters. Here is the brief. Start a Lite mode project.

    • Read your context and instructions
    • Compiled the autonomous supervisor

    I have initialized the supervisor for the CSV export project and established the success criteria based on your brief.

    Project Initialized: CSV Export Feature

    I have created the .supervisor/ directory and defined the roadmap.

    Project Objective: Implement a CLI-based CSV export utility that supports piped input and character escaping.

    IDCriterionStatus
    AC1Correctly parse piped stdin into rowsPENDING
    AC2Escape special characters (e.g., pipes, quotes)PENDING
    AC3CLI supports --output flagPENDING

    Initial Roadmap:

    • T1: Core CSV parsing logic (csvmd/core.py) - Maps to AC1, AC2
    • T2: Pipe handling and stdin integration (csvmd/__main__.py) - Maps to AC1
    • T3: CLI Argument parsing and output redirection - Maps to AC3
    
    # Supervisor State Summary
    $ python3 scripts/supervisor.py status
    Objective: CSV export feature with pipe support
    Tasks: 3 pending
    Criteria: 3 unverified
    Mode: Lite
    

    Next steps

    1. Implement the core parsing logic in csvmd/core.py.
    2. Create unit tests in tests/test_core.py to satisfy AC2.
    3. Verify T1 and update the supervisor state with the test evidence.

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

    About this skill

    Autonomous Supervisor gives an AI coding agent the working habits of a senior lead. You hand over a brief; it owns the delivery.

    What it does

    • Plans the whole project as ordered mini-goals, each with its own checkable criteria and its own gate.
    • Keeps the plan, decisions and evidence on disk, so work survives a context reset, a new session or a handoff. Chat is not memory.
    • Decides for itself. Work waits for you only for four things: spending money, irreversible actions, access it cannot get, or a change of goal. Everything else it decides, records and reports.
    • Never stops while something can still move. A blocked task does not pause the rest.
    • Treats "done" as a claim. A task or criterion is only verified with evidence the agent produced itself, and the completion gate decides "complete", not the agent's opinion.
    • Bounds failure: three failed attempts force a new strategy, three strategies force an escalation.
    • Guards parallel work: no two agents own the same file, no task starts before its dependencies are verified.
    • Audits old or interrupted projects: finds the branch with the latest work, recovers what was actually asked from past sessions, and checks each requirement against the code as it is today.
    • Works in sandboxed, read-only shells through a journal mode, so the guards and the gate stay on.

    What is inside One skill file, one zero-dependency Python script, five reference guides and eight templates. MIT licensed.

    What is measured, and what is not

    • 117 unit tests on the tool; 15 with/without evaluation cases.
    • In those cases the skill loads when it should and stays quiet on typos and quick questions.
    • On small single-session tasks a strong model scores the same without it, and the skill costs roughly 30-50% more. Its value is aimed at long, multi-session and multi-agent work, and that benefit has not yet been measured on a large project.

    Requires Python 3 and filesystem access. Not for single edits or quick questions. Subagent spawning depends on the host environment's capabilities.

    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 4 days ago

    Requires

    Python
    Git

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Environment Variables
    Listed5 days ago
    Updated4 days ago

    What's inside

    Frequently Asked Questions

    Unlike long-context prompts that inevitably drift or hallucinate progress, this skill uses a local `.supervisor/` directory to store a 'hard-coded' state machine, ensuring the agent never forgets requirements or repeats already-verified work.

    The skill is designed for any AI environment with filesystem access and shell execution, making it a perfect fit for tools like Claude Code, Cursor, and custom Python-based agentic workflows.

    You receive the Python 3 CLI tool, a suite of Markdown templates for structured project tracking, and the integration logic required to let your agent manage the supervisor independently.

    Yes, the framework includes resumption logic that audits your filesystem upon startup to determine exactly where a project left off, even if the previous chat session was completely cleared.

    The supervisor enforces 'Evidence-Based Verification,' meaning the agent cannot mark a task as complete without providing specific command outputs or file observations that prove the criteria have been met.

    Installation is straightforward: ensure you have a Python 3 environment, install the provided CLI, and initialize the supervisor in your project root using the `$SUP init` command.

    New in Agents & Orchestration