Self-hosted alternatives to review-agent-setup

R

review-agent-setup

Configure human-in-the-loop gating for AI agent review actions in Claude Code

review-agent-setup is an agent skill from the wshobson/agents collection for setting up review-agent-governance, a plugin that puts human approval in front of the review actions an AI agent takes in Claude Code. It applies to projects where an agent may post pull request reviews, comment on issues or pull requests, merge pull requests, create releases, edit CI configuration, or push to protected branches such as main, master, release or production. Each of those actions is gated behind explicit human approval, and every attempt, whether approved or denied, produces an Ed25519-signed receipt, so there is a record of what was requested and decided. The plugin's default policy does not gate posts to external notification services such as Slack or Discord, so a rule has to be added for those commands if they should be gated too. The skill walks through the setup in a project: installing the plugin, reviewing the default policy, and adding rules for commands the team wants to control. It is aimed at teams that let agents participate in code review and release work and want a clear human decision at each sensitive step. Policies decide what is blocked, so they should be reviewed by the people responsible for the repository before the agent is given access. The skill is installed with the skills CLI from the wshobson/agents repository. Its skills.sh page showed about 5.0K installs, with Gen Agent Trust Hub and Socket passing and Snyk showing a warning at the time of the scan.