AI Agent Hub
Back to plugins
⚙️

AgentsGitFlowController

Workflow Updated 2026.09.01

Run the following command in DeepSeek Harness:

dsh plugin install FeatureAgents/AgentsGitFlowController

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install FeatureAgents/AgentsGitFlowController in the DeepSeek Harness terminal to install; open-source address: https://github.com/FeatureAgents/AgentsGitFlowController

About this plugin

AI coding agents are increasingly operating directly inside your repository, yet GitFlow remains for them nothing more than a soft reminder in a project instructions file. Models skip the merge flow, push straight into protected branches, or treat human-only archive merges like any ordinary operation. AgentsGitFlowController turns the verbal agreement of following the process into a hard gate: every git command is classified and checked against your branch-role configuration before it executes, and violations are denied on the spot with a reason and a suggested next step.

The core mechanism is a pure-local, config-driven branch-role guard. A single JSON file maps the integration, preview, production, and archive roles to concrete branch names or regexes and specifies allowed update modes. The plugin intercepts at the tool-call pipeline level and does not depend on any remote service. Every denial is appended to a user-level audit log shared across worktrees, keeping the repository clean. It works with nine clients including Claude Code, Codex, Cursor, and DeepSeek Harness, so you do not need to write separate rules per agent.

Ideal for solo developers or small teams with a defined branch flow whose agents have already cut corners (direct pushes to develop, PR-less merges), and for enterprises with multi-environment pipelines that can simply add roles to the config. Not suitable for trunk-based single-branch workflows or teams that have not assigned any role to their branches.

Use Cases

  • Agent pushes directly to develop or merges a feature without a PR
  • Multi-environment pipelines require role-based merge isolation
  • Agent bypasses the defined merge flow during async collaboration

Best For

  • Solo developers with a defined branch flow
  • Small teams of 3 to 10 using AI coding agents
  • Enterprise teams running multi-environment pipelines