← Back to Skills Marketplace
swit001

Worldloops

by Josh Lee · GitHub ↗ · v0.3.0 · MIT-0
cross-platform ⚠ pending
212
Downloads
1
Stars
0
Active Installs
5
Versions
Install in OpenClaw
/install worldloops
Description
Executable world layer for OpenClaw that detects open loops, proposes governed transitions, and keeps agent execution safe with externalWrite:false.
README (SKILL.md)

WorldLoops for OpenClaw

Executable world layer for OpenClaw that detects open loops, proposes governed transitions, and keeps agent execution safe with externalWrite: false.

Most agents answer from a snapshot. WorldLoops manages open loops as state.

WorldLoops is signal-first and designed to inspect work signals without waiting for a user prompt. It surfaces governed open-loop proposals for human-in-the-loop review. Safe-by-default means proposal is not execution, approval is not external write, and externalWrite: false is the default posture.

Use this skill when the user wants to detect open loops from scattered work signals and convert them into governed proposal candidates for stateful open-loop management.

WorldLoops is not a chatbot, task bot, or uncontrolled automation trigger.

It is a governed world-model layer for interpreting work signals safely — with an auditable runtime direction in v0.3.0.

Safety boundary

WorldLoops does not send messages. WorldLoops does not send emails. WorldLoops does not create calendar events. WorldLoops does not modify external systems.

externalWrite remains false.

Runtime invocation

When the user asks you to run this skill, do the following:

  1. Change into the skill directory: {baseDir}
  2. Run the command with --silent
  3. Use provided local payload files when available
  4. If no payload files are provided, use the included fixture files
  5. Return the JSON result directly
  6. Do not modify external systems

Safe default command from the skill directory:

npm run --silent brief:reconcile --
--gmail-event scripts/fixtures/openclaw-gmail-webhook.json
--calendar-event scripts/fixtures/openclaw-calendar-events.json
--gog-gmail scripts/fixtures/gog-gmail-messages.json
--gog-calendar scripts/fixtures/gog-calendar-events.json
--message-read scripts/fixtures/openclaw-message-read.json

Notification Preferences

Runtime MVP — v0.2.7

WorldLoops stores notification preferences locally at .worldloops/notification_prefs.json. Users control time, frequency, severity thresholds, and quiet hours.

Users configure the entrypoints. WorldLoops does not auto-install background scheduling.

npm run notifications:init     # create default prefs file
npm run notifications:show     # print current prefs as JSON
npm run notifications:set -- dailyBrief.time 09:00
npm run notifications:set -- proactiveDiscovery.minSeverity high
npm run notifications:set -- quietHours.enabled true

Proactive Discovery

Runtime MVP — v0.2.7

discovery:run reads notification preferences and surfaces only candidates that match the configured severity threshold. It applies duplicate suppression using .worldloops/notification_state.json.

npm run discovery:run -- --gmail-event scripts/fixtures/openclaw-gmail-webhook.json

High-severity candidates can be surfaced before the scheduled brief. Quiet hours suppress non-critical notifications. No external writes. Discovery proposes — it does not act.

Scheduled Daily Brief

Runtime MVP — v0.2.7

brief:daily reads notification preferences, respects quiet hours, and produces a daily brief output as JSON with safety.externalWrite=false.

npm run brief:daily -- \
  --gmail-event scripts/fixtures/openclaw-gmail-webhook.json \
  --calendar-event scripts/fixtures/openclaw-calendar-events.json

This release does not auto-install cron or launchd. Users may connect brief:daily to a scheduler manually if they choose.

A daily brief includes important open loops, uncertain threads, upcoming meetings, required decisions, and signals since the last brief. The brief is a proposal, not an execution. No external writes. User approval is required for any resulting action.

Default API

By default, WorldLoops uses:

https://api.worldloops.ai

You do not need to set WORLDLOOPS_API_BASE_URL for the default demo flow.

To use a different backend, set:

WORLDLOOPS_API_BASE_URL=https://your-worldloops-api.example.com

Optional:

WORLDLOOPS_API_KEY=your_api_key

Output

The skill returns safe JSON containing:

  • brief
  • openLoops
  • proposalCandidates
  • safety.externalWrite
  • mode
  • source
  • metadata

Updating

Existing users can update with:

clawhub update worldloops

Force reinstall:

clawhub install worldloops --force

Important rules

Do not invent missing source data. Do not send messages. Do not send emails. Do not create drafts. Do not create or update calendar events. Do not write to Slack, Gmail, Calendar, GitHub, SMS, or push channels. Do not implement background daemons. Do not auto-install cron or launchd.

Return the JSON result directly.

Capability Tags
cryptorequires-oauth-tokenrequires-sensitive-credentials
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install worldloops
  3. After installation, invoke the skill by name or use /worldloops
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.3.0
WorldLoops v0.3.0 introduces the "transition receipt" system to improve auditable runtime tracking. - Added support for transition receipts with new types, storage, and scripts. - Implemented receipt listing functionality for auditable tracking of world transitions. - Added tests and example data for transition receipts. - Updated documentation to reflect new auditable runtime and receipt features. - No changes to APIs or safety boundaries; externalWrite remains false.
v0.2.8
- Updated skill metadata to include new tags for open-loop management and stateful workflows. - Clarified that WorldLoops manages open loops as state, not just from snapshots. - Added note about an upcoming auditable runtime direction (v0.3.0). - Minor documentation and description improvements for clarity and future direction. - No functional or API changes in this release.
v0.2.7
v0.2.7 removes Vercel URL exposure, adds local notification preference runtime, daily brief and discovery entrypoints, duplicate suppression, and keeps smoke validation local and safe.
v0.2.6
v0.2.6 improves Re/Fwd thread detection, proactive discovery and daily brief direction, local smoke validation, and OpenClaw/ClawHub metadata discoverability.
v0.2.5
ClawHub publish candidate with default api.worldloops.ai, gog Gmail thread detail support, OpenClaw Slack message-read envelope support, live Gmail/Calendar/Slack validation, and externalWrite:false safety boundary.
Metadata
Slug worldloops
Version 0.3.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 5
Frequently Asked Questions

What is Worldloops?

Executable world layer for OpenClaw that detects open loops, proposes governed transitions, and keeps agent execution safe with externalWrite:false. It is an AI Agent Skill for Claude Code / OpenClaw, with 212 downloads so far.

How do I install Worldloops?

Run "/install worldloops" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Worldloops free?

Yes, Worldloops is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Worldloops support?

Worldloops is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Worldloops?

It is built and maintained by Josh Lee (@swit001); the current version is v0.3.0.

💬 Comments