npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@yaniv-tg/babysitter-codex

v0.1.5

Published

Babysitter orchestration plugin for OpenAI Codex CLI - multi-step AI workflows with full upstream process library, 15 slash commands, and BOM-safe SKILL installation

Readme

babysitter-codex

Babysitter orchestration plugin for OpenAI Codex CLI. It adds structured multi-step AI workflows with quality convergence, lifecycle hooks, 15 orchestration modes, and the full upstream Babysitter process library.

This project was created by Babysitter already running on Codex.

What You Get

  • External Babysitter orchestration integrated into Codex
  • Iterative execution loop (run:create -> run:iterate -> task:post)
  • Breakpoint support for interactive approvals
  • Yolo mode for non-interactive auto-approval
  • Compatibility mode for SDK builds that expose only core run/task commands
  • Deterministic per-run trace log at <runDir>/run-trace.jsonl

Latest Version

  • Current release: 0.1.5 (2026-03-11)
  • Key additions in 0.1.5:
    • team-pinned install flow (babysitter:team-install)
    • lockfile (babysitter.lock.json)
    • layered rules resolver
    • process index cache
    • content integrity manifest + verification
    • mapping contract CI gates
    • install hardening: strip UTF-8 BOM from copied SKILL.md files to prevent Codex frontmatter parse failures

Version Control Documentation

Full Test Scenarios

  • Core scenario runner: npm run test:scenario
  • Long-session strict scenario: npm run test:long-scenario
  • Scenario runbook: test/FULL_SESSION_SCENARIO.md

The long-session scenario validates:

  • 3+ interview breakpoints
  • at least one breakpoint with 4 questions
  • generated artifact reflects selected interview choices (colors/text)
  • simulated 60-minute AI workload
  • strict 100/100 score gate

Bundled Upstream Library

babysitter-codex now bundles upstream Babysitter assets under:

  • upstream/babysitter/skills/babysit/process (full process library)
  • upstream/babysitter/skills/babysit/reference (reference docs)
  • upstream/babysitter/commands (upstream command docs)

Default process-library root for Codex mapping:

  • upstream/babysitter/skills/babysit/process

Override with:

  • BABYSITTER_PROCESS_LIBRARY_ROOT=<path>

Important: No Native /babysitter Slash Commands

Codex does not have built-in /babysitter:* commands.

Babysitter is external and activated by this skill using natural-language triggers (for example: babysitter, orchestrate, yolo, resume, doctor).

Requirements

  • Node.js 18+
  • npm 9+
  • OpenAI Codex CLI (installed and usable from terminal)

babysitter-codex installs the Babysitter SDK dependency automatically. Users do not need a separate manual SDK install for normal usage.

npm Package Name

Install from the scoped package:

  • @yaniv-tg/babysitter-codex

The unscoped babysitter-codex name is currently a security placeholder on npm and is not used for this project.

Installation (Windows)

1. Verify prerequisites (PowerShell)

node -v
npm -v
codex --version

2. Install globally from npm

npm.cmd install -g @yaniv-tg/babysitter-codex

3. Or install from local repo clone

cd C:\path\to\babysitter-codex
npm.cmd install -g .

4. Verify install

npm.cmd ls -g @yaniv-tg/babysitter-codex --depth=0

5. Restart Codex

After install, restart Codex so it loads the updated skill files.

Windows notes

  • If PowerShell blocks npx, use npx.cmd.
  • If global npm install fails with permissions, run terminal as Administrator.

Installation (macOS)

1. Verify prerequisites

node -v
npm -v
codex --version

2. Install globally from npm

npm install -g @yaniv-tg/babysitter-codex

3. Or install from local repo clone

cd /path/to/babysitter-codex
npm install -g .

4. Verify install

npm ls -g @yaniv-tg/babysitter-codex --depth=0
ls -l ~/.codex/skills/babysitter-codex/.codex/hooks/*.sh

5. Restart Codex

After install, restart Codex so it loads the updated skill files.

macOS notes

  • If shell scripts lose executable bits, rerun install or apply: chmod +x ~/.codex/skills/babysitter-codex/.codex/hooks/*.sh
  • If npx is not on PATH inside Codex, install Node via nvm and relaunch shell.

Installation (Linux)

1. Verify prerequisites

node -v
npm -v
codex --version

2. Install globally from npm

npm install -g @yaniv-tg/babysitter-codex

3. Or install from local repo clone

cd /path/to/babysitter-codex
npm install -g .

4. Verify install

npm ls -g @yaniv-tg/babysitter-codex --depth=0

5. Restart Codex

After install, restart Codex so it loads the updated skill files.

Linux notes

  • If global install needs elevated rights, use sudo npm install -g @yaniv-tg/babysitter-codex.
  • Prefer using nvm or user-owned Node install to avoid sudo where possible.
  • The installer now sanitizes UTF-8 BOM in SKILL.md during copy, preventing missing YAML frontmatter delimited by --- warnings from Codex skill loading.

Uninstall

Windows

npm.cmd uninstall -g @yaniv-tg/babysitter-codex

Linux

npm uninstall -g @yaniv-tg/babysitter-codex

macOS

npm uninstall -g @yaniv-tg/babysitter-codex

Quick Start (How To Actually Run It)

In Codex chat, use prompts like:

babysitter help
babysitter call implement authentication with tests
babysitter yolo fix lint and failing tests
babysitter resume latest incomplete run
babysitter doctor current run

You can also run the SDK CLI directly in terminal:

babysitter run:create --process-id <id> --entry <path>#<export> --inputs <file> --json
babysitter run:iterate <runDir> --json --iteration 1
babysitter task:list <runDir> --pending --json
babysitter task:post <runDir> <effectId> --status ok --value tasks/<effectId>/output.json --json

All Modes And How To Activate Them

| Mode | Example prompt to activate | What it does | |------|----------------------------|--------------| | call | babysitter call build auth with tests | Start an orchestration run (interactive) | | yolo | babysitter yolo fix failing tests end-to-end | Fully autonomous, no breakpoints | | resume | babysitter resume latest run | Resume an existing run | | plan | babysitter plan migration workflow | Plan a workflow without executing | | forever | babysitter forever monitor build health every hour | Never-ending periodic run | | doctor | babysitter doctor run 01ABC... | Diagnose run health | | observe | babysitter observe current workspace | Launch observer dashboard | | retrospect | babysitter retrospect latest run | Analyze a run and propose process improvements | | model | babysitter model set execute=gpt-5 | Set or view model routing policy | | issue | babysitter issue 123 --repo owner/repo | Start workflow from a GitHub issue | | help | babysitter help | Help and documentation | | project-install | babysitter project-install this repo | Set up a project for babysitting | | team-install | babysitter team-install | Install team-pinned runtime/content setup | | user-install | babysitter user-install for backend workflows | Set up your user profile | | assimilate | babysitter assimilate https://github.com/org/method | Assimilate external methodology |

Codex Compatibility Modes

babysitter-codex supports two runtime modes automatically:

  • full mode: SDK exposes advanced commands (session:*, profile:*, skill:*, health)
  • compat-core mode: SDK exposes core orchestration commands only (run:*, task:*, version)

In compat-core, orchestration continues and unavailable advanced commands are skipped gracefully.

Feature Flags

Advanced capabilities (event streaming, policy engine, model routing, telemetry, etc.) are gated by feature flags.

  • File-based flags: .a5c/config/features.json
  • Env-based overrides: BABYSITTER_FEATURE_<FLAG_NAME>
  • Defaults: .codex/feature-flags.js

Requested Features (1-10) Status

The top requested Codex harness features are now implemented in runtime:

  1. Session management UX: alias/tag/search resume selectors (recent, tag:<x>, search:<q>, list, name <alias>, tag +/-<x>).
  2. First-class notifications + event stream: stable v1 JSON events with id, seq, runId, webhook/slack/desktop/file sinks.
  3. Long-task autonomous mode + approvals: staged approval policy, strict allowlists, retry backoff+jitter.
  4. Multi-repo orchestration: alias/scope-based workspace routing from .a5c/workspace/repos.json.
  5. Mid-session model switching: persisted .a5c/config/model-policy.json + per-step phase routing.
  6. Plan/Act hardening: immutable plan-mode execution blocks for mutating task kinds.
  7. Richer hooks lifecycle: on-tool-error, on-policy-block, on-retry + configurable payload transforms.
  8. MCP reliability toolkit: doctor mcp diagnostics for transport/auth/path/config readiness.
  9. Native GitHub issue/PR workflow: issue fetch + comments + optional apply mode + PR comment/create helpers.
  10. Cost/token telemetry + budgets: per-iteration telemetry history, soft/hard budget phases, adaptive prompt shrinking.

New Config Files

  • .a5c/config/model-policy.json
  • .a5c/config/policies.json
  • .a5c/config/hook-transforms.json
  • .a5c/workspace/repos.json

SDK Contracts (Codex-Suitable)

  • Core required commands:
    • run:create
    • run:iterate <runDir>
    • run:status <runDir>
    • task:list <runDir>
    • task:post <runDir> <effectId>
  • Optional advanced commands:
    • session:*
    • profile:*
    • skill:*
    • health

If advanced commands are unavailable, the harness degrades to compat-core and continues orchestration with core run/task operations.

Troubleshooting

  • task:post path errors:
    • Use run-relative references like tasks/<effectId>/output.json.
  • Missing advanced SDK commands:
    • Expected in compat-core; orchestration still works with core run/task commands.

Blame Beni

CRITICAL: BENI_EXCEPTION_UNHANDLED (0xB3N1-v0.1.5)
Message: "we should keep it simple" emitted 12 milliseconds before adding lockfiles, process indexes, and 7,798 upstream assets.

Stack trace:
  at Beni.Init("tiny patch")
  at Beni.Upgrade("runtime/content split")
  at Beni.Upgrade("team install")
  at Beni.Upgrade("integrity + signatures")
  at Beni.Upgrade("15 modes")
  at PagerDuty.Trigger("prod confidence increased, sleep decreased")

Incident response:
- Severity: SEV-BENI
- Blast radius: every repository Beni can spell
- Mitigation status: containment failed, architecture improved

Repro:
1. Ask Beni for one small codex integration tweak.
2. Go make coffee.
3. Return to find lockfiles, CI contract checks, team installers, and a manifesto.

If anything breaks, open an issue and always blame Beni.
This policy is deterministic, backward-compatible, and thoroughly validated.

License

MIT