🦞 WorldLoops — Agent Execution Guard

AI agents can answer from a snapshot.

WorldLoops tracks what remains unresolved.

It turns real work signals into governed open loops — detecting unfinished responsibility, classifying severity, proposing the next transition, adjudicating whether approval is required, recording decisions, and committing local state transitions with receipts.

externalWrite:false is preserved throughout.

WorldLoops is an execution guard for AI agents — not a todo list.

Architecture

OpenClaw reads signals. WorldLoops guards execution.

WorldLoops does not need to own every connector. If a host agent can read a signal, it can pass it to Agent Execution Guard.

OpenClaw observes and interprets source signals (Gmail, Calendar, Slack, GitHub) into ObservedSignal[]. WorldLoops reads those OpenClaw-authored interpreted observations and adjudicates their lifecycle state — into active open loops, attached context, suppression receipts, and transitions. WorldLoops does not connect to Gmail, Calendar, or Slack directly. externalWrite:false is preserved throughout.

OpenClaw (reads Gmail, Calendar, Slack, GitHub)
    ↓
already-read payload
    ↓
Agent Execution Guard (WorldLoops)
    ↓
governed open loop → proposal → approval → local transition → receipt

OpenClaw Signal Handoff

OpenClaw reads. Agent Execution Guard governs.

No connectors added. No OAuth added. No external write.

A host agent (OpenClaw, gog, or any other) places an already-read payload into .worldloops/inbox/. Agent Execution Guard consumes it locally and produces a governed receipt.

.worldloops/inbox/openclaw-gmail-live.json      ← host agent places this
    ↓
npm run guard:gmail -- --input .worldloops/inbox/openclaw-gmail-live.json --compact
    ↓
Agent Execution Guard
    ↓
governed open loop → proposal → receipt
externalWrite:false

Accepted local payload formats

Agent Execution Guard can consume local payloads in these forms:

AdapterSignal JSON — fully normalized signal with source, sourceType, text, observedAt, externalWrite:false
OpenClaw-style handoff payloads — already-normalized payloads from OpenClaw host agents
gog-style Gmail payloads — { "messages": [...] } output from gog Gmail reads
gog-style Calendar payloads — { "events": [...] } output from gog Calendar reads
Slack host/plugin payloads — { "channel": "...", "messages": [...] } output from Slack host tools

gog and OpenClaw read Gmail, Calendar, and Slack. Agent Execution Guard only consumes the local JSON output they produce. No Gmail, Calendar, or Slack API call is made by WorldLoops. externalWrite:false is preserved throughout.

Supported handoff paths:

.worldloops/inbox/openclaw-gmail-live.json
.worldloops/inbox/openclaw-calendar-live.json
.worldloops/inbox/openclaw-slack-live.json
.worldloops/inbox/openclaw-github-live.json

Redacted payload examples: examples/handoff/

OpenClaw + WorldLoops

OpenClaw is excellent at observing possible signals from user-facing queries such as "What did I miss?" or "What should I do today?"

WorldLoops starts after observation.

It takes OpenClaw-observed candidate signals and adjudicates whether they are real open loops, suppressing noise such as promotional messages, FYI-only updates, duplicates, and weak evidence. Accepted signals become stateful open loops that can move through To Do, Doing, Done, Snoozed, or Escalated.

OpenClaw observes candidate signals
    ↓
OpenClaw Observation Intake (WorldLoops)
    ↓
adjudicate: accepted | suppressed | attached_context | needs_review | state_transition
    ↓
accepted → open loop (todo)
state_transition → existing loop updated (done | escalated | snoozed)
suppressed → receipt saved, no open loop created
    ↓
Morning Brief: loop lifecycle summary
externalWrite:false

Run

npm run openclaw:intake -- --input scripts/fixtures/openclaw-signal-intake/mixed-observations.json

Example output

OpenClaw observed 14 candidate signals.

WorldLoops adjudication:
- 3 accepted as new open loops
- 2 state transitions applied
- 6 suppressed as noise / no-action / promotional
- 2 attached as related context
- 1 needs review

Open loops created:
- Review MCP deck for LG partnership by Friday
- Follow up with David Kim on MCP proposal (closed_by_new_evidence)
- Confirm invoice extension request with Google Collections (escalated_due_to_deadline)

Morning Brief:
- 1 loop still open
- 1 loop closed by new evidence
- 1 loop escalated
- 6 observed signals suppressed as noise

externalWrite:false

Adjudication verdicts

Verdict	Meaning
`accepted`	Real open loop — created in `.worldloops/open_loop_states.json`
`suppressed`	Noise — promotional, FYI-only, duplicate, or weak evidence
`attached_context`	Travel event or supporting document — linked as context, no new loop
`needs_review`	Low-confidence signal — flagged for human review
`state_transition`	New evidence updates an existing loop's status

OpenClaw Observation format

{
  "id": "obs-001",
  "source": "gmail",
  "sourceId": "gmail-mcp-followup-001",
  "observedBy": "openclaw",
  "userQuery": "What did I miss?",
  "observationIntent": "new_loop",
  "title": "Follow up with David Kim on MCP proposal",
  "text": "Can you send me the updated specs by Friday?",
  "timestamp": "2026-05-22T09:15:00Z",
  "actor": "david.kim@example.com",
  "dueAt": "2026-05-24T17:00:00Z",
  "evidence": { "subject": "Re: MCP Proposal", "snippet": "..." },
  "confidence": 0.92,
  "relatedContext": null
}

📋 Daily Brief

Get one compact Agent Execution Guard summary from all three local handoff sources.

OpenClaw/gog/host tools read Gmail, Calendar, and Slack. Agent Execution Guard reads only the local payload files they produce. No Gmail, Calendar, or Slack API call is made by WorldLoops. externalWrite:false is preserved throughout.

Non-English action phrases such as Korean review requests are supported by the detector.

Expected payload locations

.worldloops/inbox/openclaw-gmail-live.json
.worldloops/inbox/openclaw-calendar-live.json
.worldloops/inbox/openclaw-slack-live.json

Run

npm run guard:daily
npm run brief:daily

Default schedule: 09:00 local time. Default delivery channel: local.

Example output — payloads connected

🦞 Agent Execution Guard Daily Brief

Sources:
✅ Gmail
✅ Calendar
✅ Slack

Open loops:

⚠️ Gmail — Review requested
From: Test Reviewer <reviewer@example.com>
Subject: Please review the submitted document
Why: review request detected
Evidence: "Please review the submitted document and send updates by EOD."
Action: Review the submitted document or reply if needed
Adjudication: requires_approval

📅 Calendar — Important context
Event: Flight to Seoul (KE 24)
When: May 21, 12:40 PM local time
Location: SFO
Reason: travel event detected, no action proposed

💬 Slack — Action requested
From: Dana
Channel: #product
Why: review or approval request detected
Evidence: "Can you review this before release?"
Action: Review and comment
Adjudication: requires_approval

✅ Safe
externalWrite:false
No email, draft, calendar event, Slack message, or external change made.

Daily Brief schedule: 09:00 local time — Delivery channel: local

Example output — Gmail no-action with samples

Promotional emails (airline offers, newsletters, discount campaigns, rewards updates) and messages containing "no action required" are automatically suppressed. They appear as no-action with an informational reason:

📧 Gmail — No actionable loop detected
Checked: 2 messages
Sample:
- From: promotions@airline.example.com / Subject: Earn double miles this weekend — promotion
- From: noreply@rewards.example.com / Subject: Your rewards balance: 12,450 miles
Reason: promotional or informational message; no reply, approval, review, deadline, or follow-up request detected

Flight, airport, and travel calendar events appear as important context without requiring approval:

📅 Calendar — Important context
Event: Departure to ICN — Korean Air KE 24
When: May 21, 12:40 PM local time
Location: SFO Terminal 2, Gate G1
Reason: travel event detected, no action proposed

Example output — Slack not connected

⬜ Slack — not connected
Reason: no Slack payload found
Next: configure OpenClaw channels.slack, then save payload to:
.worldloops/inbox/openclaw-slack-live.json

Example output — payloads not connected yet

🦞 Agent Execution Guard Daily Brief

No local handoff payloads found yet.

Add payloads here:
- Gmail: .worldloops/inbox/openclaw-gmail-live.json
- Calendar: .worldloops/inbox/openclaw-calendar-live.json
- Slack: .worldloops/inbox/openclaw-slack-live.json

Then run:
npm run guard:daily

Source systems stay untouched.
externalWrite:false

Daily Brief schedule: 09:00 local time — Delivery channel: local

Preferences

npm run brief:preferences
npm run brief:preferences:set -- --time 08:30
npm run brief:preferences:set -- --channel telegram
npm run brief:preferences:set -- --channel local

Daily Brief delivery channels include local, Telegram, Slack, Discord, SMS, and email.

Delivery

npm run brief:deliver
npm run brief:deliver -- --dry-run
npm run brief:deliver -- --channel telegram

Actual delivery to remote channels (Telegram, Slack, Discord, SMS, email) requires a host scheduler or integration. WorldLoops does not install cron, launchd, or background daemons. A host scheduler (e.g. OpenClaw) may call npm run brief:deliver at the configured time.

If no integration is active, the command exits 0 with a delivery-ready message.

If payload files are missing, the brief shows short onboarding instructions with expected file paths.

🚀 Quick Start

clawhub install worldloops
cd ~/.openclaw/workspace/skills/worldloops
npm run demo

Optional Safety Check

npm run doctor

Expected result:

🦞 Agent Execution Guard

🚨 High — Gmail callback requested
State: open

Proposal:
Review claim context and decide whether to call back or prepare a written response. This is a local planning action only — do not initiate any call, email, or external communication without an explicit decision.

Adjudication:
requires_approval

✅ Safe
externalWrite:false
No email, draft, call, or external change made.

For the broader open-loop showcase:

npm run wow

🧭 What WorldLoops does

WorldLoops does not just list tasks.

It turns real work signals into governed open loops:

detects unfinished responsibility
classifies severity
proposes the next transition
adjudicates whether approval is required
records user decisions
commits local state transitions
creates receipts
preserves externalWrite:false

🔁 From signal to governed transition

Signal
  → Open Loop
    → Severity
      → Proposal
        → Adjudication
          → User Approval
            → Local Transition
              → Receipt
                → externalWrite:false

Example real signal:

A Gmail message says:

"Please give me a call back. It is important that we discuss your injuries and this incident. Claim No. 26-99-554236."

WorldLoops produces:

🚨 High — Claim contact request
State: needs_response
Proposal: prepare callback or written response plan
Adjudication: requires_approval
Boundary: local_proposal_only
Safety: externalWrite:false

Messenger-friendly output for Telegram, Slack, Discord, WhatsApp, SMS, and mobile chat:

npm run brief:messenger -- --adapter-signal examples/adapters/gmail-claim-contact-request.example.json

After user approval:

Decision receipt created.
Loop transitioned locally.
No email sent.
No call made.
No external system changed.

This is the key product difference.

Normal assistants summarize what they see. WorldLoops tracks what remains unresolved, governs what should happen next, and records every safe transition.

🚨 Severity-aware open loops

WorldLoops classifies detected open loops by severity:

Critical / High → escalated immediately, requires approval
Medium → surfaced for review, proposal generated
Low → tracked but not escalated

High-severity loops (like a legal claim follow-up) trigger proposals with adjudication: requires_approval.

🧑‍⚖️ Proposals, adjudication, and approval

When an open loop requires action, WorldLoops creates a proposal:

what action is proposed
why it is required
what checks should be performed first
whether approval is required (requiredReview: true)
what boundary applies (local_proposal_only, read_only, etc.)

No proposal executes automatically. Human approval is required before any local transition is committed.

🧾 Receipts and audit trail

Every approved decision creates a receipt:

transition receipt (records the loop state change)
proposal decision receipt (records the approval or rejection)

Receipts are verifiable with:

npm run receipts:verify
npm run state:check

🛡 Safety boundary

WorldLoops does not send emails. WorldLoops does not post chat messages. WorldLoops does not create calendar events. WorldLoops does not modify project tools. WorldLoops does not silently change external systems.

By default:

externalWrite:false

Proposal is not execution. Approval is not external write. Plan is not execution. Contract is not external write.

🧰 Useful commands

Default demo (Agent Execution Guard compact):

npm run demo
npm run guard:demo

Daily Brief (all local inbox sources):

npm run guard:daily
npm run brief:daily
npm run brief:preferences
npm run brief:preferences:set -- --time 08:30
npm run brief:preferences:set -- --channel telegram
npm run brief:deliver
npm run brief:deliver -- --dry-run
npm run brief:deliver -- --channel telegram

Broader open-loop showcase:

npm run wow

Doctor and state:

npm run doctor
npm run doctor:mobile
npm run state:check
npm run receipts:verify

Developer tools:

npm run wow:developer
npm run loop:list
npm run proposal:list

Messenger-friendly adapter output:

npm run brief:messenger -- --adapter-signal examples/adapters/gmail-claim-contact-request.example.json

For governed adapter invocation (already-read OpenClaw payloads):

npm run guard:adapter -- --source gmail --input examples/adapters/openclaw-gmail-claim.json
npm run guard:adapter -- --source gmail --input examples/adapters/openclaw-gmail-claim.json --compact
npm run guard:gmail -- --input examples/adapters/openclaw-gmail-claim.json
npm run guard:calendar -- --input examples/adapters/openclaw-calendar-prep.json
npm run guard:slack -- --input examples/adapters/openclaw-slack-review-request.json
npm run guard:github -- --input examples/adapters/openclaw-github-pr-review.json

For adapter developers:

npm run adapter:validate -- examples/adapters/slack-message.json
npm run adapter:test -- examples/adapters/slack-message.json
npm run brief:reconcile -- --adapter-signal examples/adapters/gmail-claim-contact-request.example.json

🧑‍💻 For developers

WorldLoops is a local, safe-by-default world state layer for agent execution.

Signals become loops. Loops become proposals. Proposals become decisions. Decisions become plans. Plans become contracts. Execution remains governed.

📚 Advanced docs

Adapter guide: ADAPTER_GUIDE.md
Community adapter guide: CONTRIBUTING.md
Release history: GitHub Releases
Changelog: CHANGELOG.md

🚫 What WorldLoops is not

WorldLoops is not:

a chatbot
a todo app
a Zapier clone
an uncontrolled automation runner
a tool that lets agents freely write to external systems
a replacement for human judgment

WorldLoops keeps humans in control without making them the bottleneck.

📱 Telegram demo wrapper

src/scripts/telegramTestBot.ts is a local demo wrapper for testing WorldLoops adjudication over Telegram. It consumes local OpenClaw-observed JSON payloads and does not connect to Gmail, Slack, Calendar, GitHub, or any external API.

OpenClaw observes. WorldLoops adjudicates. This wrapper bridges the two for live demo.

Input priority

When /brief is sent, the bot selects its input in this order:

.worldloops/inbox/openclaw-observations.json — mode: inbox-openclaw-observations
.worldloops/inbox/telegram-observations.json — mode: inbox-telegram-observations
scripts/fixtures/openclaw-signal-intake/mixed-observations.json — mode: demo-fixture

Setup

If you have an OpenClaw gateway bot using the same Telegram token, stop it first to avoid a polling conflict (409).

# Seed the demo inbox with the mixed-observations fixture
npm run telegram:seed-demo

# Start the bot
npm run telegram:test

Commands

Command	Description
`/help`	List all commands and natural language examples
`/status`	Bot version and status
`/source`	Which input file would be used and whether it exists
`/brief`	Run WorldLoops adjudication and show open loops
`/worldloops`	Same as `/brief`

Natural language also triggers /brief:

"오늘 내가 할 일이 뭐야?"
"뭐 빠진 거 없어?"
"어제 열린 루프 중 닫힌 거 있어?"

Sample payload

scripts/fixtures/openclaw-signal-intake/demo-observations.json is a small 3-signal sample showing one actionable loop, one follow-up, and one suppressed promotional email.

🔗 Links

Website: https://worldloops.ai
API: https://api.worldloops.ai
ClawHub: worldloops

Give agents a world, not just tools.

Close the loop. Keep the world safe.

Name		Name	Last commit message	Last commit date
.github/PULL_REQUEST_TEMPLATE		.github/PULL_REQUEST_TEMPLATE
app		app
dist		dist
docs/assets		docs/assets
examples		examples
schemas		schemas
scripts		scripts
src		src
tests		tests
.clawhubignore		.clawhubignore
.gitignore		.gitignore
ADAPTER_GUIDE.md		ADAPTER_GUIDE.md
CHANGELOG.md		CHANGELOG.md
CONTRIBUTING.md		CONTRIBUTING.md
README.md		README.md
SKILL.md		SKILL.md
index.html		index.html
package-lock.json		package-lock.json
package.json		package.json
tsconfig.json		tsconfig.json

Latest commit

History

Folders and files

Repository files navigation