Spacedock plugin
Spacedock splits agent work into stages with decision gates, so a human approves or returns each result together with evidence.
- 85 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Agents work in git worktrees and can change project files; a decision can be delegated to an agent without a human.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/spacedockDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Agents work in git worktrees and can change project files; a decision can be delegated to an agent without a human.
This is third-party code. Review the repository files before installing.
What it does
Spacedock works inside Claude Code, Codex or Pi as a plugin and a set of skills. Work is split into stages, each with a stated quality bar, and a decision gate with a stage report closes each one: findings, verdicts, artifacts, anomalies. Review runs as a separate stage with fresh context and no access to the maker reasoning. Stages that change shared state run in their own git worktree. The `/spacedock:survey`, `/spacedock:commission` and `/spacedock:debrief` commands describe the current process, create a workflow from plain text files and capture session learnings.
Who it is for. Developers running a dozen agent sessions at once and drowning in ship-or-not questions without a plan or a definition of done.
Good fit when
- Tasks are queued and decisions need to be taken in a batch, not one session at a time.
- A project was vibe-coded and you need to see how the work actually went.
- Agent output is checked by a separate reviewer without access to its reasoning.
- Decisions must be stored with evidence to trace a bad outcome later.
Not a fit when
- The task is a single file and stages with gates only add work.
- You need a fully autonomous run with no human involved.
- None of Claude Code, Codex or Pi is in your workflow.
Example request
Study the agent session history in this project and show which process I use and which decisions are still waiting on me.Limitations
The project is early, as the authors say in the README. It needs a coding agent harness, and the main install path goes through Homebrew. Parallel stages create git worktrees, so check disk space and repository state. Decisions can be delegated to an agent, which removes direct human control over approval. Stage decisions are recorded together with evidence, so reports may contain code and log fragments: store them as carefully as the repository itself.
How to disable. Remove the plugin from Claude Code or Codex, then run `brew uninstall spacedock`. Delete workflow directories and worktrees by hand.
FAQ
Which harnesses does Spacedock support?
Claude Code, Codex and Pi. Through skill systems it also runs in most other harnesses.
How do I run it in Codex or Pi?
Swap the subcommand: `spacedock codex "/spacedock:survey"` or `spacedock pi "/spacedock:survey"`.
Is there a channel with early builds?
Yes, the edge channel. It is chosen at install time: the `spacedock@next` cask or `SPACEDOCK_CHANNEL=edge` for the script.
What does `/spacedock:debrief` do?
It captures session learnings so the next session starts from them.
What does `/spacedock:survey` do?
It reads your agent session history and shows the process you use, how you close work and which decisions are still open.
Discussion
No questions or reviews yet.
Related
PluginsMulti-agent orchestration for Claude Code: teams, roles, autopilot and task routing across skills and subagents
A Claude Code plugin that hands coding tasks to Codex agents, monitors them and verifies the results independently
Claude Code plugin that runs complex tasks as step-by-step workflows with quality gates and reviews changes against project rules.
A plugin that syncs Claude Code, Codex CLI and Antigravity on one project and lets them hand work to each other.
Claude Code supervises while Codex, opencode and pi work on your quota: delegated edits, independent review, model discussions
An operating layer for Claude Code and Codex: one /do entry point, project memory across sessions, safety hooks, cost telemetry and parallel agents
README in short
An orchestrator for Claude Code, Codex and Pi: stages with criteria, gate reports and review in fresh context. Installed through Homebrew, and the first launch sets up the plugin itself.