Opencode Delegate

    1

    Claude Code hands big, low-judgment grunt work (tests, docstrings, boilerplate, fixtures) to free opencode models, then reviews the diff itself.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    opencode-delegate

    Example session with this skill installed

    Add Google-style docstrings to every public function in src/textkit/, with Args and Returns sections. Tests must still pass. Use the free model.

    • Read your context and instructions
    • Compiled the opencode-delegate

    Claude sizes the job with git ls-files and wc -l, then runs:

    git status --porcelain && oc-delegate -c "pytest -q" "Goal: add a Google-style docstring ..." && git diff --stat

    model: opencode/deepseek-v4.1-flash-free
    changed:
    M src/textkit/strings.py
    M src/textkit/numbers.py
    ...
    check: passed

    Claude reads the diff once, fixes anything wrong, and reports: "The free model added the docstrings; the tests pass and I checked the diff."

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

    About this skill

    A Claude Code skill that hands low-judgment grunt work to free opencode models, so your Claude usage goes to the parts that need it.

    Claude stays in charge. It decides what to delegate, writes the prompt, runs the free model through a sandboxed worker (oc-delegate), then reviews the diff and runs the tests itself.

    What the skill teaches Claude

    • When delegation pays off (roughly 150+ lines of new content or hand edits across several files) and when it doesn't (small jobs, one-line sed jobs, design, tricky bugs, security).
    • A four-call workflow: size the job with git ls-files/wc -l, delegate with the project's test command as a check, read the diff once, fix only what's wrong.
    • How to handle model fallback, failing checks, and files outside git diff.

    The wrapper and worker (installed from GitHub)

    • oc-delegate runs a chain of free models. If one fails or hangs, only the files it edited are reverted and the next is tried. Your own uncommitted changes are left alone.
    • With -c "<test command>", failing checks go back to the worker to fix, up to two retries.
    • The cc-worker agent can only read and edit files in the repo: no shell, no web, no .git, .env*, keys or credential files. On Linux with bwrap, writes are sandboxed to the repo.

    In the benchmark on the project's fixture, a 33-docstring job cost $0.31 of Claude usage with the skill vs. $0.63 on average without it.

    Setup: this listing is the skill text. The oc-delegate wrapper and cc-worker agent come from the GitHub repo: clone it and run ./install.sh. See the README for the full install, including a prompt you can paste into Claude Code to have it install everything.

    • Small jobs cost more Claude usage delegated than done directly; the skill tells Claude to skip them.
    • Free OpenCode Zen models may log prompts and use them for training. Don't use it in confidential repos.
    • The worker's deny list matches file names only; a secret inside config.yaml is still readable.
    • The worker has no shell, so it can't run tests itself; the wrapper's -c check and Claude do the verifying.
    • The free model lineup changes; the default chain may need updating via OC_DELEGATE_MODELS.
    • Works only inside a git repo.

    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

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed today

    • Free with an Agensi account

    Requires

    Claude Code
    opencode CLI
    bash, git, jq
    GNU coreutils
    oc-delegate (from GitHub repo)
    Git

    Permissions

    Allowed Hosts

    opencode.ai (OpenCode Zen free models)

    File Scopes

    current
    git
    repo
    (worker
    reads
    and
    edits
    files)
    ~/.local/state/oc-delegate.log

    Repo files the worker reads are sent to the free model provider. The worker cannot read .git, .env*, *secret*, *credentials*, *.pem, *.key or ssh keys.

    Listedtoday

    What's inside

    Frequently Asked Questions

    Claude Code hands big, low-judgment grunt work (tests, docstrings, boilerplate, fixtures) to free opencode models, then reviews the diff itself.

    This skill uses the open SKILL.md standard. Use it through an MCP-compatible agent connected to Agensi, and check the compatibility information on this page for confirmed agents and required capabilities.

    Connect your AI agent to Agensi through MCP, then ask it to use opencode-delegate on your task. Your agent loads the latest approved version each time. Paid skills require a purchase or active subscription. Skill files are not available for download; your licence does not allow you to save, copy or export any part of a skill.

    New in LLM Ops