Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    handOff.

    1

    Save and resume agent tasks across sessions and tools using structured handoff files and automated log recovery.

    $10

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    handOff.

    handOff.

    Example session with this skill installed

    My tokens are almost gone and I need to switch to Cursor. Save my progress on the user authentication refactor so the next agent knows where I stopped and what to do next.

    • Read your context and instructions
    • Compiled the handoff

    I have captured the current state of the authentication refactor and updated HANDOFF.md at the project root. I also ran the log parser to include the last five terminal commands and the specific trace from the failing unit test.

    HANDOFF: User authentication refactor

    Updated: 2024-10-24 14:30 UTC | By: claude-3-5-sonnet | Status: In progress

    Next agent: read this whole file, check it against the repo (git status), then continue from "Next steps". Keep this file updated as you work.

    Goal

    Refactor the AuthService to support JWT rotation and move sensitive logic from AuthController to the service layer. Done means all tests in tests/auth.spec.ts pass and the .env.example is updated.

    Where we stopped

    Mid-refactor of src/services/auth.service.ts. I just extracted the generateRefreshToken method, but src/controllers/auth.controller.ts is currently broken because it still expects the old method signature. The last test run failed with a TypeError at line 42.

    Next steps

    1. Update AuthController.login to use the new AuthService.createSession signature.
    2. Fix the broken import in src/middleware/session.ts.
    3. Run npm test tests/auth.spec.ts to verify the fix.

    Done so far

    • Created src/types/auth.d.ts for standardized token shapes.
    • Moved signToken logic from controller to service.

    Tried and failed / gotchas

    • Attempted to use crypto.randomUUID() for tokens, but the environment requires uuid v4 for compatibility with the existing Redis schema.

    Key files

    • src/services/auth.service.ts - Core logic being refactored.
    • src/controllers/auth.controller.ts - Currently in a broken state.

    Next steps

    • Open this project in Cursor.
    • Run git status to confirm the uncommitted changes in the service layer.
    • Type "Read HANDOFF.md and continue" to resume.
    Read HANDOFF.md

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

    About this skill

    Stop losing progress when context windows fill up or usage limits hit without warning. When a session ends abruptly, the next agent starts from zero, wasting time and tokens re-learning the task. HANDOFF solves this by maintaining a HANDOFF.md file at the project root that captures the goal, the plan, failed attempts, and the exact state of half-finished work. It ensures continuity across different tools, models, and sessions.

    What it does

    • State preservation captures goals, specific code locations, and the current hypothesis in a format readable by any LLM.
    • Log recovery extracts data from session logs of major coding CLIs (Claude Code, Cursor, Aider, etc.) to rebuild state after a crash.
    • Automated context tracking runs a Python script to pull recent commands, changed files, and error messages directly into the handoff file.
    • Cross-tool compatibility allows you to switch from Claude to Gemini or Cursor while keeping the agent focused on the next immediate step.
    • Session protection monitors active sessions to warn about impending limits and ensures a checkpoint exists before the context window expires.

    How it works

    1. Initialize by asking the agent to "write a handoff" or "save progress" when tokens are low or a milestone is reached.
    2. Execute script using python3 scripts/handoff.py now --tool AGENT to automatically populate repo facts and session history.
    3. Resume work in a new session by instructing the agent to "read HANDOFF.md and continue."
    4. Recover logs by running the sessions command to find and digest the history of a session that ended due to a timeout or limit.

    Frameworks & tools

    The core logic resides in a standalone Python 3.8+ script using only the standard library and git. It supports session log parsing for Claude Code, Codex, Cursor, Gemini CLI, Copilot CLI, OpenCode, Qwen Code, Codebuff, and Aider.

    Why this beats prompting it yourself

    Manual prompting often forgets the "why" behind decisions or fails to list the specific files that are currently broken. This skill enforces a structured template and uses a script to pull objective data (like exact error messages and git status) that a human or LLM might misremember when context is tight.

    Use cases

    • Migrating a complex task from one AI tool to another due to rate limits.
    • Resuming a long-running refactor after a weekend or session timeout.
    • Recovering context after a CLI tool crashes or the API returns a credit error.
    • Checking in progress during multi-step features to prevent hallucinations as context fills.

    Known limitations

    Requires Python 3.8+ for log parsing and automated repo tracking. Background session protection requires a local installation to run the watcher process.

    Requires Python 3.8+ for log parsing. Automated recovery depends on the specific CLI tool storing logs on disk. Background watcher requires manual installation via the install command.

    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
    Git

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Environment Variables

    Writes only HANDOFF.md in the project, and only rewrites the lines between its own marker comments. Agents' session logs and databases are read-only (databases are opened in read-only mode). No network calls, no packages installed, secrets masked before writing. Two optional commands go further, and the skill runs them only after the user says yes: "install" sets up a background watcher (a login item and a "handoff" command) that refreshes HANDOFF.md after every step, and "skill install" copies the skill into the user's other coding CLIs' skills folders. Both have matching uninstall commands.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    HANDOFF creates a structured `HANDOFF.md` file that captures the current goal, specific code locations, and failure logs, preventing agents from "re-learning" the task or hallucinating when starting a new session.

    The skill includes a Python-based recovery script that can parse local session logs from tools like Claude Code, Aider, and Cursor to rebuild the project state even if the session crashed without a manual save.

    Yes, the skill is designed for cross-tool compatibility; you can generate a handoff in one environment (e.g., Claude Code) and have a different model (e.g., Gemini or GPT-4o) resume immediately from the exact same point.

    You receive the core Python automation script (compatible with Python 3.8+), the structured handoff templates, and the session watcher tool for proactive checkpointing.

    While the basic handoff template is manual, the included Python script automates the heavy lifting by pulling git status, recent error messages, and command history directly into the file for you.

    It is a lightweight standalone tool requiring only Python 3.8+ and Git, making it easy to integrate into any project folder without complex dependencies or heavy installation processes.

    New in Agents & Orchestration