Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    handoff-kit: session pickup & handoff contract.

    1

    Standardized AI handoff protocol with structured journals, state-of-play docs, and automated Python validation.

    Free

    1 useSecurity scanned
    handoff-kit: session pickup & handoff contract.

    handoff-kit: session pickup & handoff contract.

    Example session with this skill installed

    That's it for today. Wrap up the session: record what we did in the journal, update the handoff, and make sure it validates.

    • Read your context and instructions
    • Compiled the handoff-kit

    Session closed. The journal has a new entry, the handoff is updated in place, and the validator passes.

    What changed in the record

    FileChange
    JOURNAL.mdPrepended session-2026-10-21-release-v1 after the <!-- SESSION ENTRIES --> marker. Older entries untouched.
    docs/state-of-play.mdEvidence baseline moved to 2e400e6 (tag v1.0.0). Finished items removed, new next action set. 779 of 800 words.
    .gitignoreAdded dist/.

    Journal entry

    <a id="session-2026-10-21-release-v1"></a>
    

    2026-10-21 Released v1.0.0

    Objective and constraints

    Merge the skill package, tag it, and produce the release zip. Listing it is the user's step.

    Changes and decisions

    PR #2 rebase-merged to main as 2e400e6 (linear history kept), tagged v1.0.0, tag pushed. The zip is built from the tag, not the working tree:
    git -c core.autocrlf=false archive --format=zip --prefix=handoff-kit/ -o dist/handoff-kit-1.0.0.zip v1.0.0:skills/handoff-kit

    Validation and evidence

    The check, kit and GitGuardian checks passed on PR #2 at 3b7b233. The merged tree is identical to the tested tree. The zip holds 14 files with LF endings; extracted to a fresh directory, it installed into an empty project that passed the validator.
    Not verified: marketplace review, and a plugin install.

    State at the end

    main and v1.0.0 pushed. Downstream repositories untouched.

    Next action and completion condition

    Install the plugin and run a pickup in this repository.
    Completion condition: the skill triggers from an install, and the listing result is recorded.

    Do not repeat

    Build release zips from the tag with core.autocrlf=false. Zipping a Windows checkout ships CRLF files.

    Validator

    $ python tools/check-handoff.py
    Handoff structure, links and anchors OK.
    This checks structure, not truth: read the evidence and the contradictions yourself.
    

    Status, scoped

    • VERIFIED: CI on the PR, and a clean install from the release zip.
    • NOT VERIFIED: the marketplace review and a plugin install. They are the next action, not assumed done.

    Next session starts with

    Read docs/state-of-play.md, then the entry above. The first job is the plugin install check.

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

    About this skill

    AI sessions often end with fragmented context and lost momentum. This skill implements a structured handoff protocol that ensures every session starts exactly where the last one left off. It maintains a single source of truth for project state, preventing the "context reset" tax that slows down multi-session AI development.

    What it does

    • Install handoff infrastructure to set up a standard AGENTS.md contract, state-of-play document, and dated journal.
    • Automate session pickups by parsing existing state, git status, and recent commits to synchronize context instantly.
    • Formalize session ends by generating dated journal entries and updating the live state-of-play document with new evidence baselines.
    • Validate handoff integrity using a dependency-free Python script that checks for broken links, word counts, and stale markers.
    • Synchronize project tools by configuring AGENTS.md specifically for tools like Claude, Cursor, and GitHub Copilot.

    How it works

    1. Install the kit into a repo using install.py to create the standard file structure and config.
    2. Pickup work by reading docs/state-of-play.md and confirming the current objective before starting tasks.
    3. Journal progress by prepending anchored entries to JOURNAL.md using the provided session templates.
    4. Validate the state using python tools/check-handoff.py to ensure all links and status markers are compliant.

    Frameworks & tools

    This skill is a pure process implementation using Python 3.8+ standard library scripts. It integrates with Git for state tracking and is compatible with all major LLMs and AI coding assistants including Claude, Gemini, Cursor, and Copilot.

    Why this beats prompting it yourself

    Generic prompts for "summarizing what we did" result in hallucinated progress or vague bullet points. This skill enforces a schema-based validation via a Python script, ensuring the handoff actually links to evidence and adheres to hard word limits for context window efficiency.

    Use cases

    • Resuming a complex refactor after a 24-hour break without re-explaining the plan.
    • Handing off a repository from a local dev session to a remote agent.
    • Maintaining a readable audit trail of AI-generated changes for human review.
    • Ensuring CI/CD pipelines fail if the AI forgets to update the project state.

    Known limitations

    The validator enforces document structure and link integrity but cannot verify the factual accuracy of the AI's reported progress. Requires Python 3.8+ to run the validation scripts.

    Validator checks structure, links, and limits only, not factual truth. Requires Python 3.8+. Existing AGENTS.md files are never overwritten automatically.

    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

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 26 days ago

    • Free with an Agensi account

    Requires

    Python 3.8+
    Git (optional, for the --base CI check)
    Python
    Git

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access

    File Scopes

    AGENTS.md
    JOURNAL.md
    handoff.config.json
    docs/state-of-play.md
    docs/session-template.md
    tools/check-handoff.py
    CLAUDE.md
    GEMINI.md
    .cursorrules
    .github/copilot-instructions.md
    .github/workflows/handoff.yml

    The installer never overwrites an existing file unless --force is passed, and --dry-run shows the plan first. No network calls, no credentials, no telemetry.

    Listed26 days ago

    What's inside

    Frequently Asked Questions

    This kit solves 'context drift' by replacing vague AI summaries with a validated schema, including a 'state-of-play' doc and a dated journal, ensuring your AI starts every session with a perfect memory of previous progress.

    The Handoff Kit is compatible with any LLM or IDE, including Claude, Cursor, GitHub Copilot, and Gemini, as it relies on standardized Markdown files and a Python validation script rather than platform-specific code.

    You receive the core `AGENTS.md` contract, standardized templates for state-of-play and journaling, and a dependency-free Python validation script to ensure your documentation remains structurally sound.

    Yes, the included `check-handoff.py` script is designed to run in any Python 3.8+ environment, allowing you to automatically block merges or fail builds if the AI fails to update the session state correctly.

    Setup is nearly instantaneous; simply run the `install.py` script within your repository to generate the necessary file structure and configuration tailored for your AI tools.

    No. Existing files are skipped and listed; --force is the only way to overwrite. If you already have a journal, add the marker and one anchored entry by hand.

    No. It checks structure, dates, word limits, links and anchors. Whether the content is accurate still needs someone to read the evidence. The kit says so plainly.

    Yes. Everything enforced lives in handoff.config.json.

    Ask your agent to "upgrade the handoff validator", or run install.py <repo> --upgrade. It replaces only the validator and CI workflow, and reports local differences without touching them.

    New in Workflow Automation