Workshop with a review pair
Optional five-seat Workshop variant with two independent reviewers, one Claude Code and one Codex. The four-seat Workshop remains the recommended default. Both install as workshop in the same directory; choose one variant.
Listed by OpenRig maintainers · Reviewed for listing on 8 Oct 2026 at commit ec063e3d1124 · not a security audit
github.com/mvschwarz/openrig-world · rigs/workshop-review-pair · commit ec063e3d1124
The team
How much OpenRig has tested it
For the recommended configuration, from OpenRig's published test status.
- Harnesses tested by OpenRig on macOS and LinuxClaude Code 2.1.290 and Codex 0.160.0 on macOS, OpenRig 0.6.6, 6 Oct 2026; Claude Code 2.1.299 (updated to 2.1.302) and Codex 0.163.0 on Linux, OpenRig 0.6.11, 18 Oct 2026OpenRig checked each harness on its own, but hasn't run this team as a whole at commit ec063e3d1124, or on OpenRig 0.7.20.
Harness checks
What OpenRig ran for each harness: the basis of the test status above.
- Claude Code on macOSPassedClaude Code 2.1.290 · OpenRig 0.6.6 · 6 Oct 2026
- Claude Code on LinuxPassedClaude Code 2.1.299 (updated to 2.1.302) · OpenRig 0.6.11 · 18 Oct 2026
- Codex on macOSPassedCodex 0.160.0 · OpenRig 0.6.6 · 6 Oct 2026
- Codex on LinuxPassedCodex 0.163.0 · OpenRig 0.6.11 · 18 Oct 2026
What it will do
Which agents start
- orch.lead on codex, profile lead, working folder "."
- dev.build on claude-code, profile builder, working folder "."
- dev.qa on claude-code, profile qa, working folder "."
- review.r1 on claude-code, profile code-reviewer, working folder "."
- review.r2 on codex, profile code-reviewer, working folder "."
What they can do
- orch.lead: Can run shell commands as the launching user, subject to runtime and host policy. Permission prompts: off. Codex runs with full access and never asks. The bundle sets OpenRig's built-in yolo policy.
- dev.build: Can run shell commands as the launching user, subject to runtime and host policy. Permission prompts: off. Claude bypasses permissions. The bundle sets OpenRig's built-in yolo policy.
- dev.qa: Can run shell commands as the launching user, subject to runtime and host policy. Permission prompts: off. Claude bypasses permissions. The bundle sets OpenRig's built-in yolo policy.
- review.r1: Can run shell commands as the launching user, subject to runtime and host policy. Permission prompts: off. Claude bypasses permissions. The bundle sets OpenRig's built-in yolo policy.
- review.r2: Can run shell commands as the launching user, subject to runtime and host policy. Permission prompts: off. Codex runs with full access and never asks. The bundle sets OpenRig's built-in yolo policy.
What they are told
- agents/workshop/guidance/lead.md (orch.lead)
- agents/workshop/skills/agent-operated-software (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/agent-operated-workflows (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/applying-a-permission-policy (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/claude-compaction-restore (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/delegating-work (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/forming-an-openrig-mental-model (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/loading-addressable-markdown (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/messaging-the-human (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/mission-slice-sop (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/openrig-operating-model (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/openrig-user (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/orienting-to-an-inherited-seat (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/bundle-skills/ponytail (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/queue-handoff (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/refocusing (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/retiring-and-inheriting-a-seat (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/seat-continuity-and-handover (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/skills/session-compaction-and-restore (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- agents/workshop/bundle-skills/using-the-tui-together (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- CULTURE.md (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- startup/context.md (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- startup/lead-first-move.md (orch.lead)
- startup/review-pair-lead.md (orch.lead)
- agents/workshop/guidance/builder.md (dev.build)
- startup/review-pair-builder.md (dev.build)
- agents/workshop/guidance/qa.md (dev.qa)
- agents/workshop/guidance/code-reviewer.md (review.r1, review.r2)
- startup/review-pair-reviewer.md (review.r1, review.r2)
What else runs
- proof the agent read its startup instructions: authenticated, on after_files; fresh_start, restore (orch.lead, dev.build, dev.qa, review.r1, review.r2)
- runtime add-ons (settings, hooks, MCP servers): agents/workshop/runtime/claude-activity-hooks.json, on if selected runtime supports and loads this resource (dev.build, dev.qa, review.r1)
What it writes
Each depends on your configured destinations, what the runtime supports, and files already in place when it installs or launches.
- AGENTS.md in the seat's working folder: add or update OpenRig's section (orch.lead, review.r2)
- .agents/skills/agent-operated-software in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/agent-operated-workflows in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/applying-a-permission-policy in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/claude-compaction-restore in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/delegating-work in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/forming-an-openrig-mental-model in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/loading-addressable-markdown in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/messaging-the-human in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/mission-slice-sop in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/openrig-operating-model in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/openrig-user in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/orienting-to-an-inherited-seat in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/ponytail in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/queue-handoff in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/refocusing in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/retiring-and-inheriting-a-seat in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/seat-continuity-and-handover in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/session-compaction-and-restore in the seat's working folder: project (orch.lead, review.r2)
- .agents/skills/using-the-tui-together in the seat's working folder: project (orch.lead, review.r2)
- CLAUDE.local.md in the seat's working folder: add or update OpenRig's section (dev.build, dev.qa, review.r1)
- .claude/skills/agent-operated-software in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/agent-operated-workflows in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/applying-a-permission-policy in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/claude-compaction-restore in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/delegating-work in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/forming-an-openrig-mental-model in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/loading-addressable-markdown in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/messaging-the-human in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/mission-slice-sop in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/openrig-operating-model in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/openrig-user in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/orienting-to-an-inherited-seat in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/ponytail in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/queue-handoff in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/refocusing in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/retiring-and-inheriting-a-seat in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/seat-continuity-and-handover in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/session-compaction-and-restore in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/skills/using-the-tui-together in the seat's working folder: project (dev.build, dev.qa, review.r1)
- .claude/settings.local.json in the seat's working folder: merge settings or managed hooks (dev.build, dev.qa, review.r1)
- . in the install folder: unpack the team's files
Outside addresses
- agentskills.io
- code.claude.com
- github.com
- learn.chatgpt.com
- www.apache.org
- www.openrig.dev
What you need
- codex (runtime: orch.lead, review.r2)
- codex: the user's own runtime account or provider configuration (login: orch.lead, review.r2)
- claude-code (runtime: dev.build, dev.qa, review.r1)
- claude-code: the user's own runtime account or provider configuration (login: dev.build, dev.qa, review.r1)
- OpenRig daemon >=0.6.6 (version)
- OpenRig CLI >=0.6.6 (version)
Not known before launch
- host and native settings: Effective permissions, login, model availability, runtime versions, host catalogs/plugins and configured paths are not checked before launch.
- runtime behavior: Scripts and agents can compute other actions and addresses. Literal URL occurrences are not a prediction of network traffic.
- built-in and host integration: The installed OpenRig/runtime adds culture, onboarding, session identity and host context, and can provision runtime hook/settings files. Those bytes and effects are outside this archive.
Install
Copy the whole block. Its last line fetches exactly this configuration from GitHub and starts it.
This team declares broad access. Its Codex and Claude Code seats default to non-interruptive launches: OpenRig accepts supported harness warnings with launch flags. Sign-in and unsupported notices may still need you. Use --no-non-interruptive to opt out.
Before you run it, you need:
From the team's author:
- Install and run the rig as your normal user, not root: Claude Code refuses to bypass permissions as root or under sudo.
- Open a shell at the root of your clone of the repository you want to work on, set up the way its README says. Optionally add OpenRig's public context pack, then run the install from that same shell.
- Keep OpenRig's files out of your commits: from the root of your clone, add CLAUDE.local.md, AGENTS.md, /.openrig/, .claude/settings.local.json and gate-lane-verdict.json to .git/info/exclude (the README has the command). The lead adds any that are missing when it starts.
- If you ask Claude Code to do this setup for you, plan to run some steps yourself. In OpenRig's tests on macOS, Claude Code's default auto mode declined the rig up line ([Create Unsafe Agents], because it starts new agents) on 2.1.289 and 2.1.290, and in an earlier 2.1.289 run also declined adding a permission rule for itself ([Self-Modification]; add any rule yourself with /permissions or in .claude/settings.local.json) and rig context add --git ([Untrusted Code Integration]), even with an allow rule for that command. Run rig context add --git and rig up yourself in a terminal, from the root of your clone.
- This team declares broad access and non-interruptive launches: Claude bypasses permissions, Codex runs unsandboxed and never asks, and Pi gets full resource trust. OpenRig accepts supported Claude/Codex warning screens with launch flags; sign-in and unsupported harness notices still require you.
- codex (runtime: orch.lead, review.r2)
- codex: the user's own runtime account or provider configuration (login: orch.lead, review.r2)
- claude-code (runtime: dev.build, dev.qa, review.r1)
- claude-code: the user's own runtime account or provider configuration (login: dev.build, dev.qa, review.r1)
- OpenRig daemon >=0.6.6 (version)
- OpenRig CLI >=0.6.6 (version)
rig context add --git https://github.com/mvschwarz/openrig-worldrig up https://github.com/mvschwarz/openrig-world/tree/ec063e3d11249a7e7b74d210f93d08d8c490b600/rigs/workshop-review-pair --cwd . --target ~/rigs/workshopIt installs these three. The install prints the same three, so you can check they match:
- Commit
ec063e3d1124- Configuration
dev.build=claude-code,dev.qa=claude-code,orch.lead=codex,review.r1=claude-code,review.r2=codex- Package digest
60e9e59ba529Covers the bundle's files, not your host's settings.
It unpacks the team's files into ~/rigs/workshop, outside your project. Your project gets only the seats' own files.
npx skills add https://github.com/mvschwarz/openrig/tree/v0.7.20/skills/rigs --skill rigs --agent claude-codeRun it in your project folder, then type /rigs in Claude Code. Give it this team's page and tell it what you want to work on.
npx skills add https://github.com/mvschwarz/openrig/tree/v0.7.20/skills/rigs --skill rigs --agent codexRun it in your project folder, then type $rigs in Codex. Give it this team's page and tell it what you want to work on.
Needs Node.js 22 or 24 (on Apple silicon, 22), which provides npx. If your shell says npx isn't found, Node isn't installed or isn't on your PATH. How to install Node.js →
After you run it
Tell us how it went. Community reports appear on this page, apart from OpenRig's own testing.