jj-flow: Multi-Repo Orchestration for Coding Agents

    1

    - **Feature Lifecycle Automation** managing the ANALYZE-PLAN-DELIVER-ACCEPT-ARCHIVE loop with persistent state.

    $1

    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

    jj-flow: Multi-Repo Orchestration for Coding Agents

    Example session with this skill installed

    I just finished the auth-fix in Project A. Can you port it to Project B and Project C, then run a read-only review?

    • Read your context and instructions
    • Compiled the jj-flow

    I have resolved the source commit for the auth-fix in Project A and identified the target repositories from the global map. I will now establish the baseline for Project B and Project C to begin the precise porting process.

    Multi-Repo Handoff & Review Status

    ProjectActionStatusArtifact
    Project ASource Delta Analysis[COMPLETED]commit-8f2a1b
    Project BFeature Porting ($jj-same)[IN_PROGRESS].workflow/ralph/task-auth-port-b/
    Project CFeature Porting ($jj-same)[PENDING]Waiting for B
    Project BRead-only Review ($jj-review)[QUEUED]ASSIGNMENT-REVIEW-B.md
    Project CRead-only Review ($jj-review)[QUEUED]ASSIGNMENT-REVIEW-C.md

    Next steps

    1. Approve Port B: Review the adaptation plan for Project B's specific architecture once generated.
    2. Verify Attestations: Check the sandbox_evidence_ref in .workflow/ for automated test results in each target.
    3. Dispatch Closeout: Once reviews are [OK], trigger /jj-dispatch to prepare the final integration PRs.

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

    About this skill

    Managing features across multiple related repositories manually is slow and prone to drift. The jj-flow package provides a project-family orchestration layer that automates the lifecycle of a feature, from initial analysis in one repo to precise porting across sibling projects.

    What it does

    • Multi-Repo Mapping indexing project families into a global map for cross-project visibility.
    • Feature Lifecycle Automation managing the ANALYZE-PLAN-DELIVER-ACCEPT-ARCHIVE loop with persistent state.
    • Precise Feature Porting syncing requirement invariants and code deltas to sibling repos without simple file copying.
    • Read-only Code Review spawning sub-agents to provide structured feedback without modifying business logic.
    • Automated Git Closeout handling Chinese Conventional Commits, branch merges, and pushes in a single pass.
    • Advanced Execution Engines utilizing swarm intelligence, fixed SDLC pipelines, or dynamic role-specs for complex tasks.

    How it works

    1. Initialize the project map using $jj-init to group related repositories and bootstrap the knowledge base.
    2. Execute a task loop with $jj-ralph to take a requirement from analysis through to a verified archive.
    3. Port the completed feature to other repositories in the family using $jj-same to handle target-specific adaptations.
    4. Review and Dispatch changes across the project family with $jj-review for audits and $jj-dispatch for coordinated releases.

    Frameworks & tools

    Built for coding agents on platforms like Claude Code, Codex, Grok, and Qoder. It utilizes Git for version control and stores orchestration state in .workflow/ directories.

    Why this beats prompting it yourself

    Standard prompts lack persistent state across repository boundaries and often fail at complex multi-step lifecycles. This skill enforces a control plane where checkpoints are advanced by verifiable artifacts and git commits rather than unreliable chat affirmations.

    Use cases

    • Finishing a feature in one repo and instantly porting the logic to mobile and web sibling repos.
    • Running an adversarial swarm search to optimize a complex algorithm before implementation.
    • Orchestrating a large-scale change across ten repositories with a central dispatch control plane.
    • Conducting structured, multi-role SDLC reviews for high-stakes production code.

    Known limitations

    Does not advance checkpoints based on verbal chat claims; requires physical artifacts or git evidence. Claude Code does not support the experimental offline evaluation skill.

    Checkpoints require git/artifact evidence, not just chat. No Claude support for jj-evaluated. Requires npx installation of @brewer/jj-flow.

    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 5 days ago

    Requires

    @brewer/jj-flow
    Node.js
    Python
    Git

    Permissions

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

    What's inside

    Frequently Asked Questions

    jj-flow solves the 'context drift' problem by creating a persistent orchestration layer across multiple repositories, ensuring that feature logic ported from one project to another remains consistent and verified through git-backed state management.

    This skill is designed for high-autonomy coding agents like Claude Code, Codex, and Grok; it requires an environment with Git access and the ability to read/write to a local `.workflow/` directory for state persistence.

    Unlike simple prompts, jj-flow enforces a 'Physical Artifact Only' rule, meaning the agent cannot mark a task as complete based on chat alone—it must detect specific git commits or file changes to advance the workflow.

    Setup is handled via the `$jj-init` command, which indexes your project family and creates a global map; once initialized, the agent uses this map to navigate dependencies and port code between sibling repos.

    Yes, jj-flow includes specialized execution engines like 'swarm intelligence' for algorithmic optimization and 'fixed SDLC pipelines' for teams that require strict adherence to professional development stages.

    New in Agents & Orchestration