Open Loop Audit
Scan your open loops (tasks, commitments, recurring work) and classify each as real delegation (work that leaves your desk) vs simulated work (work that lands back on it).
When to trigger
- User says "audit my loops", "what should I delegate", "what's simulated work", "open loop audit"
- Periodic review of task lists, agent outputs, or automation candidates
Phases
Phase 1 — Gather Open Loops
Collect open loops from all available sources:
- Obsidian vault — scan for unchecked tasks (
- [ ]), open projects, recurring notes
- Notion — check task databases and project boards if connected
- Calendar — look for recurring meetings that generate follow-up work
- Agent fleet — review scheduled tasks and their outputs (the agent platform scheduled tasks, cron jobs)
- Git — open PRs, stale branches, unresolved issues
Deduplicate and compile into a single list.
Phase 2 — Classify Each Loop
For each open loop, apply the Land or Leave test:
| Question |
Real Delegation |
Simulated Work |
| Does completing this remove work from your plate? |
Yes - it's gone |
No - it creates a report/summary you still have to read |
| Does it close a loop with an external party? |
Yes - email sent, PR merged, invoice paid |
No - internal document that sits in a folder |
| Could you forget about it after delegating? |
Yes - it's truly off your desk |
No - you'll check the output and redo parts of it |
| Does the output go to someone else? |
Yes - client, teammate, vendor |
No - it comes back to you |
Classify as:
- DELEGATE — real delegation candidate, work leaves your desk
- SIMULATED — looks productive but lands back on you
- HYBRID — partially delegatable, needs decomposition
Phase 3 — Prioritize Delegation Candidates
Rank DELEGATE items by:
- Frequency — how often does this recur?
- Time cost — how long does it take you each time?
- Agent-readiness — can an existing agent/tool handle this now?
- Risk — what's the cost of a bad delegation?
Phase 4 — Report
Output format:
## Open Loop Audit — [DATE]
### Ready to Delegate (LEAVE your desk)
- [ ] [Task] — [Agent/tool that could handle it] — [Est. time saved/week]
### Simulated Work (LANDS on your desk)
- [Task] — [Why it's simulated] — [Could it be restructured?]
### Hybrid (Needs decomposition)
- [Task] — [Delegatable part] vs [Part that stays with you]
### Quick Wins
Top 3 items to delegate this week, with specific next steps.
Verification
- Every open loop has a classification with reasoning
- At least one source was checked (vault, calendar, agents, git)
- Quick wins section has actionable next steps, not vague recommendations
Rules
- Never classify something as DELEGATE if the user will still review every output manually — that's SIMULATED
- Be honest about hybrid items. Most real work has a delegatable component and a human component.
- If no sources are accessible, ask the user to list their open loops manually
Source
Nate's Newsletter, 2026-03-28 — "The open loop audit prompt that separates real delegation from simulated work"
Framework: Does the work land on your desk or leave it?
1---2name: open-loop-audit3description: Audit open loops and classify each as real delegation candidate vs simulated work using Nate's "land or leave" framework4---56# Open Loop Audit78Scan your open loops (tasks, commitments, recurring work) and classify each as **real delegation** (work that leaves your desk) vs **simulated work** (work that lands back on it).910## When to trigger1112- User says "audit my loops", "what should I delegate", "what's simulated work", "open loop audit"13- Periodic review of task lists, agent outputs, or automation candidates1415## Phases1617### Phase 1 — Gather Open Loops1819Collect open loops from all available sources:20211. **Obsidian vault** — scan for unchecked tasks (`- [ ]`), open projects, recurring notes222. **Notion** — check task databases and project boards if connected233. **Calendar** — look for recurring meetings that generate follow-up work244. **Agent fleet** — review scheduled tasks and their outputs (the agent platform scheduled tasks, cron jobs)255. **Git** — open PRs, stale branches, unresolved issues2627Deduplicate and compile into a single list.2829### Phase 2 — Classify Each Loop3031For each open loop, apply the **Land or Leave test**:3233| Question | Real Delegation | Simulated Work |34|----------|----------------|----------------|35| Does completing this remove work from your plate? | Yes - it's gone | No - it creates a report/summary you still have to read |36| Does it close a loop with an external party? | Yes - email sent, PR merged, invoice paid | No - internal document that sits in a folder |37| Could you forget about it after delegating? | Yes - it's truly off your desk | No - you'll check the output and redo parts of it |38| Does the output go to someone else? | Yes - client, teammate, vendor | No - it comes back to you |3940Classify as:41- **DELEGATE** — real delegation candidate, work leaves your desk42- **SIMULATED** — looks productive but lands back on you43- **HYBRID** — partially delegatable, needs decomposition4445### Phase 3 — Prioritize Delegation Candidates4647Rank DELEGATE items by:481. **Frequency** — how often does this recur?492. **Time cost** — how long does it take you each time?503. **Agent-readiness** — can an existing agent/tool handle this now?514. **Risk** — what's the cost of a bad delegation?5253### Phase 4 — Report5455Output format:5657```58## Open Loop Audit — [DATE]5960### Ready to Delegate (LEAVE your desk)61- [ ] [Task] — [Agent/tool that could handle it] — [Est. time saved/week]6263### Simulated Work (LANDS on your desk)64- [Task] — [Why it's simulated] — [Could it be restructured?]6566### Hybrid (Needs decomposition)67- [Task] — [Delegatable part] vs [Part that stays with you]6869### Quick Wins70Top 3 items to delegate this week, with specific next steps.71```7273## Verification7475- Every open loop has a classification with reasoning76- At least one source was checked (vault, calendar, agents, git)77- Quick wins section has actionable next steps, not vague recommendations7879## Rules8081- Never classify something as DELEGATE if the user will still review every output manually — that's SIMULATED82- Be honest about hybrid items. Most real work has a delegatable component and a human component.83- If no sources are accessible, ask the user to list their open loops manually8485## Source8687Nate's Newsletter, 2026-03-28 — "The open loop audit prompt that separates real delegation from simulated work"88Framework: Does the work land on your desk or leave it?
m2ai-portfolio