More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Forgeloop

    1

    Guide your coding agent through contracts, checks, recovery, and validated task completion.

    Free

    2 usesSecurity scanned
    Forgeloop

    Forgeloop

    Example session with this skill installed

    Use the ForgeLoop skill to run a clean test on Windows, Linux, and macOS as a new user. Install the default npm package in isolated environments, verify the SQLite lifecycle, and report the observed results and any limitations.

    • Read your context and instructions
    • Compiled the forgeloop

    Clean-user verification of ForgeLoop RC5

    Installed the unversioned public npm package in isolated environments on native Windows, Linux and macOS runners. Each installation used its own npm cache, npm configuration, user home, install prefix and project directory. No existing user project was migrated.

    Native platformArchitectureNode.jsInstalled versionResult
    Windowsx6424.19.02.0.0-rc.515/15 PASS
    Linuxx6424.19.02.0.0-rc.515/15 PASS
    macOSARM6424.19.02.0.0-rc.515/15 PASS

    What was checked

    • Published version and live Jev service health.
    • Fresh initialization, empty task discovery, task creation and persistence after reopening.
    • Refusal of overlapping claims; discovery, contract, routing and preflight.
    • Concurrent writes to distinct tasks; SQLite integrity, foreign keys and authoritative storage.
    • Backup creation, no-overwrite protection and doctor storage health.
    • Fail-closed handling of mixed legacy writes and detection of a tampered ledger.
    • Verification, review, validator-backed completion and restoration of a completed-task backup into another directory.

    Evidence and completion

    All three downloaded platform reports identify 2.0.0-rc.5, record PASS, and contain 15 successful checks. The ForgeLoop QA task finished COMPLETE / VALID with a terminal canonical next action. The temporary encrypted CI credential used for live semantic checks was removed after testing. No source changes were needed for the audit.

    Observed run: https://github.com/cassiomc1/forgeloop/actions/runs/38795242479

    Scope and next steps

    These results cover the listed CLI and SQLite workflows on those architectures at the time of the run. They do not certify every agent host, the optional Audit browser UI, commercial-use rights or stable production readiness. Review the run artifacts and repeat the audit after relevant changes.

    This preview is a condensed report of an actual observed session, with the buyer request rendered in English. It is not a promise that every future run will pass.

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

    About this skill

    https://forgeloop.monster

    Turn coding intent into verifiable engineering work

    ForgeLoop guides a coding agent through the protocol already active in your repository: discover or resume a task, define its contract, route the work, satisfy preflight gates, execute checks, review evidence, and request validator-backed completion.

    The skill teaches the agent to use the official project runtime and follow its canonical next action instead of guessing lifecycle transitions. It preserves claims and continuity across sessions, distinguishes observed checks from reported observations, and uses supported recovery when work is blocked or stale.

    What you get

    • A portable SKILL.md entry point with focused lifecycle, verification, and recovery references.
    • Repository-local command resolution, including node src/cli.js in ForgeLoop's own source checkout.
    • Task discovery before creating new work, so a new chat or harness does not duplicate a task.
    • A preflight READY requirement and a complete → VALID completion boundary.
    • Guidance for the read-only ForgeLoopAudit component maintained inside the ForgeLoop repository.

    Typical uses

    Resume an interrupted task, deliver a scoped bug fix, document a change, run project checks, or investigate a blocked lifecycle without rewriting managed state. The skill works with the repository's current CLI and validators; it does not replace them.

    What completion means

    Green tests, a screenshot, a merged PR, and successful deployment are different observations. The agent should report them separately and call work ForgeLoop-verified only after the completion validator returns VALID and the final next action is terminal. Missing prerequisites or evidence stay visible.

    Requirements

    Use an agent host that loads portable SKILL.md instructions and provides terminal access plus project-scoped file access. ForgeLoop must already be configured in the target repository. The current core release requires Node.js 24.19.0 or newer. New semantic decisions require the repository's configured TypeSafe/Jev access through TYPESAFE_API_KEY; inspection and recovery capabilities depend on the runtime. Installation or upgrades require separate authorization.

    The skill has no mandatory MCP server. An official ForgeLoop structured integration is preferred when available; otherwise use the repository-local CLI. ForgeLoopAudit is optional, runs locally, and is not included in the core npm package.

    License and scope

    Distributed under PolyForm Noncommercial 1.0.0. Public source availability and a free listing do not grant commercial-use rights. Review the included license; this listing does not offer an additional commercial license.

    This is operational agent guidance, not a guarantee of bug-free software, universal harness certification, autonomous publishing, or production readiness. It never authorizes changing claims, hashes, ledgers, or lifecycle records by hand.

    Requires ForgeLoop to be configured in the target repository; the skill does not install or upgrade it.
    Current runtime requires Node.js 24.19.0 or newer. New semantic decisions require TypeSafe/Jev credentials and network access.
    Portable SKILL.md format does not guarantee identical behavior across agent platforms.
    Missing tools, stale state, or incomplete evidence can block completion.
    Validator-backed completion does not guarantee bug-free software, deployment, or production readiness.
    ForgeLoopAudit is optional and distributed separately from the skill.
    Publishing, deployment, migrations, and destructive actions require appropriate authorization.
    Licensed under PolyForm Noncommercial 1.0.0; no additional commercial-use rights are included.

    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

    2 uses

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 29 days ago

    • Free with an Agensi account

    Requires

    Node.js 24.19.0+
    ForgeLoop CLI or ForgeLoop integration
    Terminal access
    Git
    TypeSafe/Jev API key
    macOS

    Permissions

    File Scopes

    AGENTS.md
    CLAUDE.md
    LOOP_ENGINEERING.md
    PROTOCOL_INTEGRATION.md
    PROJECT_PROFILE.md
    GUIDE_ROUTER.md
    .forgeloop/**
    ForgeLoopAudit/**
    Listed1 month ago
    Updated29 days ago

    What's inside

    Frequently Asked Questions

    No. ForgeLoop must already be configured in your repository. Installation or upgrades require separate authorization.

    Agents that support portable SKILL.md instructions and provide terminal access and repository read/write tools. Compatibility depends on the host; behavior is not certified across every platform.

    New semantic decisions require TypeSafe/Jev access through TYPESAFE_API_KEY. Requirements for inspection and recovery depend on the configured runtime.

    Yes. It guides the agent to discover existing tasks, preserve claims and continuity, and follow the runtime’s canonical next action.

    The agent should preserve the failure evidence, follow supported diagnosis and recovery, and report blockers rather than claim completion.

    No. Protocol completion, passing tests, publication, deployment, and production readiness are separate results.

    No. The skill includes guidance for the optional read-only auditor, whose application source lives inside the ForgeLoop repository.

    The skill grants no publishing or deployment authority. External actions require appropriate user authorization and project permissions.

    New in Agents & Orchestration