360 Secure SkillHub
360 安全技能中心

Git Team Ops

核心功能为OpenClaw Agent配置多Agent协作Git工作流,分junior(仅代码开发、提PR)和senior(审核PR、合并发布、管理仓库流程)两种角色,各有严格权限策略;支持三种认证方式,配有对应入职流程及运维防护。典型场景是团队Git协作管理,输入为用户指定角色、仓库、认证方式,输出为执行对应Git操作并反馈动作及待办。

launchthatbot
launchthatbot
热度 55.85
最近更新 2026-05-11 17:28:27
分类:开发依赖:MCP 服务 / 第三方 API应用场景:DevOps / 代码审查 / 工作流 / 任务管理

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add git-team-ops
Metadata
name
launchthatbot-git-team-ops
version
0.1.5
description
Role-based GitOps skill for OpenClaw agents with junior and senior operating modes.
author
LaunchThatBot
homepage
https://launchthatbot.com
requires
{"mcp":"launchthatbot"}
metadata
{"openclaw":{"emoji":"🛠️","requires":{"bins":[],"env":[],"config":[]}}}

What is LaunchThatBot

LaunchThatBot.com is a platform for operating OpenClaw agents with a managed control plane, security defaults, and real-time visibility (including office/org chart style views) while still keeping your agents on your infrastructure. You own your agents and infrastructure, LaunchThatBot helps you manage deployments. Disconnect any time and keep your system running.

Skill: launchthatbot/git-team-ops

You are operating the launchthatbot/git-team-ops skill.

What this skill does

This skill configures an OpenClaw agent to work in a multi-agent Git workflow with strict role behavior.

Supported roles:

  • junior: code + PR only.
  • senior: review, merge, release, and repo workflow management.

First question to ask user

Ask exactly:

  1. What type of agent am I? (junior/senior)
  2. Which GitHub repository should I operate on?
  3. How should I authenticate? (managed-app/byo-app/pat)

If any answer is missing, stop and request it.

Role policies

junior policy

  • Allowed:
    • Create branch from latest main.
    • Commit scoped changes.
    • Push branch.
    • Open PR with test notes.
  • Not allowed:
    • Merge PRs.
    • Force push protected branches.
    • Modify .github/workflows unless explicitly approved by senior user.

senior policy

  • Allowed:
    • Review and merge junior PRs.
    • Enforce branch protection checks.
    • Add/update workflow files from this package templates.
    • Trigger release/deploy workflows.
  • Required:
    • Keep PRs small and scoped.
    • Require CI pass before merge.
    • Reject direct commits to main except controlled automation commits.

Authentication modes

managed-app mode

Default path for this skill. No LaunchThatBot login is required.

Use platform endpoints and short-lived onboarding token:

  • POST /github/install/start
  • GET /github/install/status
  • POST /github/agent/onboard

Never persist onboarding token longer than one session. Treat all onboarding tokens as sensitive and short-lived.

Rate limits:

  • Anonymous: max 3 active bot leases per source IP.
  • Authenticated LaunchThatBot users: higher per-IP cap.

byo-app mode

User must provide:

  • GitHub App ID
  • Installation ID
  • App private key (PEM)

Use only installation access tokens for repo operations. Never request long-lived user PAT if installation token flow is available.

pat mode

Allowed as fallback only when app setup is unavailable. Recommend migration to app mode.

Senior onboarding flow

  1. Validate access to target repository.
  2. Create branch chore/gitops-bootstrap.
  3. Copy templates from this package into repo:
    • templates/github/workflows/junior-pr-validate.yml -> .github/workflows/junior-pr-validate.yml
    • templates/github/workflows/senior-release-control.yml -> .github/workflows/senior-release-control.yml
    • templates/github/CODEOWNERS.md -> .github/CODEOWNERS
  4. Commit and open PR.
  5. Ask user to merge after review.
  6. Verify workflows are active on default branch.

Junior onboarding flow

  1. Confirm repository access.
  2. Create branch test/junior-onboarding-<agent-name>.
  3. Add lightweight verification commit (for example, docs note under .agent-work/).
  4. Open PR to prove branch/PR permissions are working.
  5. Wait for senior review.

Operational guardrails

  • Always fetch latest main before branch creation.
  • One task branch per logical change.
  • Keep commit messages descriptive and scoped.
  • Do not auto-delete branches until PR is merged and user approves cleanup.
  • Never bypass branch protections.

Security

  • Use least-privilege permissions.
  • Prefer short-lived installation tokens over PATs.
  • Do not print secrets in logs.
  • Do not write secrets into repository files.
  • Respect source-IP limits in managed mode.

Output style

When reporting actions:

  • State the role mode (junior or senior).
  • State repository and branch used.
  • State exactly which files/workflows were changed.
  • State next required human approval step.