OpenAgents Control
A Claude Code plugin with seven specialized subagents and approval gates: the agent proposes a plan and waits for confirmation before touching code
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Subagents write and change code in the project, though the plan requires explicit approval
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/openagents-controlDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Run one by one in the Claude Code chat
/plugin marketplace add darrenhinde/openagentscontrol
/plugin install oac@oac-marketplaceInstall the plugin: /plugin marketplace add darrenhinde/OpenAgentsControl, then /plugin install oac. After install, download the context files: /oac:setup --core.
Other ways from the author
/plugin marketplace add darrenhinde/OpenAgentsControl
/plugin install oac
/oac:setup --coreBeta version of the plugin.
This is third-party code. Review the repository files before installing.
What it does
OpenAgents Control (OAC) is built on top of OpenCode and adds a six-stage workflow with mandatory human approval before execution. Before generating code, the agent loads the project's user-described patterns and standards through a context system, so it doesn't output generic code that ignores decisions the team already made. Work is split across seven subagents: task-manager, context-scout, context-manager, coder-agent, test-engineer, code-reviewer and external-scout, plus nine workflow skills and six user commands. There is a Claude Code plugin version installed via a marketplace, and an OpenCode version.
Who it is for. For dev teams tired of rewriting generated code to match their conventions and who want an explicit plan-approval step.
Good fit when
- You need the agent to follow specific project patterns instead of generic code
- You need a mandatory plan-approval step before the agent starts changing files
- The task needs delegation to specialized subagents like a tester or reviewer
Not a fit when
- You need a quick one-off answer without intermediate approval gates
- The project is too small to describe context and patterns upfront
Example request
Using OAC, add a login endpoint: propose a plan first and wait for my approvalLimitations
The Claude Code plugin is marked beta. The context system's usefulness depends on how thoroughly the team describes its patterns upfront, a separate ten-to-fifteen-minute setup step.
How to disable. Remove the plugin with /plugin uninstall oac, or remove the marketplace with /plugin marketplace remove darrenhinde/OpenAgentsControl.
Security check
- Subagents write and change code in the project, though the plan requires explicit approval
README in short
The README explains the problem of generic code that ignores project patterns and proposes a fix through a context system and approved steps. It describes a Propose to Approve to Execute process with delegation to specialized subagents, separately documents installing the Claude Code plugin via a marketplace with /plugin marketplace add and /plugin install oac, and the /oac:setup command for initial context loading.
FAQ
Does it only work with Claude Code?
No, the main version is built on OpenCode; the Claude Code plugin is a separate beta variant of the same workflow.
Is plan approval mandatory?
Yes, that is a stated feature: the agent always asks for confirmation before executing rather than acting autonomously.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything