babysitter-codex
Orchestrate complex, multi-step AI workflows with quality convergence loops, event-sourced state, and human-in-the-loop approval gates. Use when the user wants to babysit a task, orchestrate a workflow, run quality-gated development, resume a previous orchestration run, diagnose run health, plan without executing, set up a project or user profile for babysitter, or assimilate an external methodology. Also use when the user mentions "babysitter", "orchestrate", "babysit", "quality loop", or "convergence loop".
Best use case
babysitter-codex is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Orchestrate complex, multi-step AI workflows with quality convergence loops, event-sourced state, and human-in-the-loop approval gates. Use when the user wants to babysit a task, orchestrate a workflow, run quality-gated development, resume a previous orchestration run, diagnose run health, plan without executing, set up a project or user profile for babysitter, or assimilate an external methodology. Also use when the user mentions "babysitter", "orchestrate", "babysit", "quality loop", or "convergence loop".
Teams using babysitter-codex should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/babysitter-codex/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How babysitter-codex Compares
| Feature / Agent | babysitter-codex | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Orchestrate complex, multi-step AI workflows with quality convergence loops, event-sourced state, and human-in-the-loop approval gates. Use when the user wants to babysit a task, orchestrate a workflow, run quality-gated development, resume a previous orchestration run, diagnose run health, plan without executing, set up a project or user profile for babysitter, or assimilate an external methodology. Also use when the user mentions "babysitter", "orchestrate", "babysit", "quality loop", or "convergence loop".
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
Related Guides
Cursor vs Codex for AI Workflows
Compare Cursor and Codex for AI coding workflows, repository assistance, debugging, refactoring, and reusable developer skills.
AI Agents for Coding
Browse AI agent skills for coding, debugging, testing, refactoring, code review, and developer workflows across Claude, Cursor, and Codex.
Best AI Skills for Claude
Explore the best AI skills for Claude and Claude Code across coding, research, workflow automation, documentation, and agent operations.
SKILL.md Source
# Babysitter for Codex CLI Orchestrate complex, multi-step workflows with event-sourced state management, hook-based extensibility, and human-in-the-loop approval gates. ## Choosing a Mode Based on the user's request, read the appropriate sub-skill from `.codex/skills/babysitter/<mode>/SKILL.md` (relative to this skill's install directory) and follow its instructions. | User intent | Mode | Sub-skill to read | |-------------|------|-------------------| | Start an orchestration run (default) | call | `call/SKILL.md` | | Run autonomously, no interaction | yolo | `yolo/SKILL.md` | | Resume an existing run | resume | `resume/SKILL.md` | | Plan a workflow without executing | plan | `plan/SKILL.md` | | Start a never-ending periodic run | forever | `forever/SKILL.md` | | Diagnose run health | doctor | `doctor/SKILL.md` | | Launch observer dashboard | observe | `observe/SKILL.md` | | Analyze previous run improvements | retrospect | `retrospect/SKILL.md` | | Set or view model routing policy | model | `model/SKILL.md` | | Work directly from a GitHub issue | issue | `issue/SKILL.md` | | Help and documentation | help | `help/SKILL.md` | | Onboard a project | project-install | `project-install/SKILL.md` | | Install team-pinned setup | team-install | `team-install/SKILL.md` | | Set up user profile | user-install | `user-install/SKILL.md` | | Assimilate external methodology | assimilate | `assimilate/SKILL.md` | If unclear, default to `call/SKILL.md`. ## SDK CLI Quick Reference The babysitter SDK CLI (`babysitter` or `npx @a5c-ai/babysitter-sdk`) drives all orchestration: ``` babysitter run:create --process-id <id> --entry <path>#<export> ... babysitter run:iterate <runDir> --json --iteration <n> babysitter run:status <runDir> --json babysitter task:list <runDir> --pending --json babysitter task:post <runDir> <effectId> --status ok --value <file> --json ``` Compatibility levels: - Core required: `run:create`, `run:iterate`, `run:status`, `task:list`, `task:post` - Optional advanced: `session:*`, `profile:*`, `skill:*`, `health` If advanced commands are missing, continue in `compat-core` mode and do not block orchestration. ### Canonical argument shapes (current Babysitter SDK) - `run:status <runDir> --json` - `task:list <runDir> --json` - `session:init --session-id <id> --state-dir .a5c --json` - `session:associate --session-id <id> --state-dir .a5c --run-id <runId> --json` - `hook:log --hook-type <type> --log-file .a5c/logs/hooks.jsonl --json` ## Result Posting Protocol 1. Write result value to `tasks/<effectId>/output.json` 2. Post: `babysitter task:post <runDir> <effectId> --status ok --value tasks/<effectId>/output.json` 3. Never write `result.json` directly -- the SDK owns that file. ## Hook-Driven Loop After each effect is posted, stop the session. The stop hook re-invokes Codex to continue the next iteration. Do not loop within a single session.
Related Skills
babysitter
Orchestrate via @babysitter. Use this skill when asked to babysit a run, orchestrate a process or whenever it is called explicitly. (babysit, babysitter, orchestrate, orchestrate a run, workflow, etc.)
babysitter:project-install
Set up a project for babysitting. Research the codebase, build project profile, install tools.
babysitter:model
Set or view model routing policy for plan/execute/review phases.
babysitter:user-install
Set up babysitter for yourself. Install deps, build user profile, configure tools.
babysitter:resume
Resume orchestrating an existing babysitter run.
babysitter:yolo
Start babysitting in non-interactive mode — no user interaction or breakpoints, fully autonomous execution.
babysitter:observe
Launch the babysitter observer dashboard for real-time run monitoring.
babysitter:retrospect
Analyze a completed or in-flight run and propose process improvements for future runs.
babysitter:assimilate
Assimilate an external methodology, harness, or specification into babysitter process definitions.
babysitter:call
Start a babysitter orchestration run. Use this command to start babysitting a complex workflow.
babysitter:doctor
Diagnose babysitter run health — journal integrity, state cache, effects, locks, sessions, logs, and disk usage.
babysitter:plan
Plan a babysitter workflow without executing it. Focus on creating the best process possible.