Back to directory

OpenCode Power Pack for Claude Code, Codex, OpenCode + Pi

A collection of 54 portable skills for Claude Code, Codex, OpenCode and Pi, covering code review, security audit, feature development, frontend design, project memory and authoring tools, plus an opt-in native sandbox for safer command execution.

OpenCode Power Pack for Claude Code, Codex, OpenCode + Pi

The skills work as portable workflows across all four supported agents, invoked with names such as /opencode-power-pack:code-review or /opencode-power-pack:feature-dev once installed. A recent addition is a lightweight, fail-closed sandbox runner: one contained command tree with no container image or daemon, giving all four agents a shared capability contract with least-privilege profiles (observe, develop, network-read or publish) and explicit escalation checks.

The sandbox uses native Seatbelt isolation on macOS and Bubblewrap on Linux, and fails closed rather than retrying outside the sandbox when isolation is unavailable, an escalation is undeclared, confirmation is missing, or the backend fails; its current guarantee covers the command and its descendants being shell-contained, not automatic host routing or whole-agent isolation.

It installs as a Claude Code plugin marketplace entry, a Codex plugin, or a global npm package that adds an opencode-power-pack CLI for running the sandbox directly.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.