Dispatch Handoff Brief Generator
Generate a structured delegation brief for handing work off to an autonomous agent (Anthropic Dispatch, the agent platform mission tasks, or any agent that works independently).
When to trigger
- User says "write a handoff brief", "delegate this to dispatch", "dispatch brief", "create a delegation brief"
- User is about to create a mission task or scheduled agent run and needs a well-structured prompt
- User says "hand this off" or "have [agent] handle this"
Phases
Phase 1 — Understand the Task
Ask (if not already clear):
- What needs to get done? (the objective)
- How will you know it worked? (success criteria)
- What tools/apps/accounts does the agent need access to?
- What should the agent do if something goes wrong?
Keep it to one short exchange. Don't over-interview.
Phase 2 — Generate Brief
Produce a structured brief in this format:
# Handoff Brief: [Task Name]
## Objective
[One sentence. What is the end state when this is done?]
## Success Criteria
- [ ] [Measurable outcome 1]
- [ ] [Measurable outcome 2]
- [ ] [Measurable outcome 3]
## Context
[2-3 sentences of background the agent needs. What does it need to know that isn't obvious from the objective?]
## Tools & Access Required
- [Tool/app 1] — [what it's used for in this task]
- [Tool/app 2] — [what it's used for in this task]
## Steps (Suggested)
1. [Step 1]
2. [Step 2]
3. [Step 3]
(Agent may deviate if it finds a better path — these are guidelines, not a script.)
## Verification
How the agent should confirm the work is done:
- [Check 1]
- [Check 2]
## Escalation Conditions
Stop and notify the human if:
- [Condition 1 — e.g., "access denied to any required tool"]
- [Condition 2 — e.g., "task will take longer than 30 minutes"]
- [Condition 3 — e.g., "any destructive or irreversible action needed"]
## Constraints
- [Time limit, if any]
- [Budget/cost limit, if any]
- [Things the agent must NOT do]
Phase 3 — Adapt to Target
If the user specifies a target agent:
- Dispatch — keep the brief as plain text, ready to paste into the Dispatch text interface
- the agent platform mission task — format as a single prompt string suitable for
the dispatch CLI.js create
- Scheduled task — add cron timing and note any state that persists between runs
- Generic — use the full markdown format above
Verification
- Brief has all 7 sections filled (Objective, Success Criteria, Context, Tools, Steps, Verification, Escalation)
- Success criteria are measurable, not vague ("email sent" not "email handled well")
- Escalation conditions include at least one safety boundary
- No section says "N/A" — if truly not applicable, remove the section
Rules
- Briefs should be self-contained. The agent receiving this should need zero follow-up questions.
- Default to over-specifying escalation conditions. Autonomous agents should fail loudly, not silently.
- Never include credentials in the brief. Reference environment variables or tool names, not actual keys/passwords.
- Keep it under 500 words. If the brief is longer, the task probably needs decomposition.
Source
Nate's Newsletter, 2026-03-28 — "The open loop audit prompt that separates real delegation from simulated work"
Pattern: Structured handoff briefs for real delegation to autonomous agents.
1---2name: dispatch-handoff-brief3description: Generate structured delegation briefs for Anthropic Dispatch or any autonomous agent handoff with objective, success criteria, tools needed, verification, and escalation conditions4---56# Dispatch Handoff Brief Generator78Generate a structured delegation brief for handing work off to an autonomous agent (Anthropic Dispatch, the agent platform mission tasks, or any agent that works independently).910## When to trigger1112- User says "write a handoff brief", "delegate this to dispatch", "dispatch brief", "create a delegation brief"13- User is about to create a mission task or scheduled agent run and needs a well-structured prompt14- User says "hand this off" or "have [agent] handle this"1516## Phases1718### Phase 1 — Understand the Task1920Ask (if not already clear):211. What needs to get done? (the objective)222. How will you know it worked? (success criteria)233. What tools/apps/accounts does the agent need access to?244. What should the agent do if something goes wrong?2526Keep it to one short exchange. Don't over-interview.2728### Phase 2 — Generate Brief2930Produce a structured brief in this format:3132```markdown33# Handoff Brief: [Task Name]3435## Objective36[One sentence. What is the end state when this is done?]3738## Success Criteria39- [ ] [Measurable outcome 1]40- [ ] [Measurable outcome 2]41- [ ] [Measurable outcome 3]4243## Context44[2-3 sentences of background the agent needs. What does it need to know that isn't obvious from the objective?]4546## Tools & Access Required47- [Tool/app 1] — [what it's used for in this task]48- [Tool/app 2] — [what it's used for in this task]4950## Steps (Suggested)511. [Step 1]522. [Step 2]533. [Step 3]54(Agent may deviate if it finds a better path — these are guidelines, not a script.)5556## Verification57How the agent should confirm the work is done:58- [Check 1]59- [Check 2]6061## Escalation Conditions62Stop and notify the human if:63- [Condition 1 — e.g., "access denied to any required tool"]64- [Condition 2 — e.g., "task will take longer than 30 minutes"]65- [Condition 3 — e.g., "any destructive or irreversible action needed"]6667## Constraints68- [Time limit, if any]69- [Budget/cost limit, if any]70- [Things the agent must NOT do]71```7273### Phase 3 — Adapt to Target7475If the user specifies a target agent:76- **Dispatch** — keep the brief as plain text, ready to paste into the Dispatch text interface77- **the agent platform mission task** — format as a single prompt string suitable for `the dispatch CLI.js create`78- **Scheduled task** — add cron timing and note any state that persists between runs79- **Generic** — use the full markdown format above8081## Verification8283- Brief has all 7 sections filled (Objective, Success Criteria, Context, Tools, Steps, Verification, Escalation)84- Success criteria are measurable, not vague ("email sent" not "email handled well")85- Escalation conditions include at least one safety boundary86- No section says "N/A" — if truly not applicable, remove the section8788## Rules8990- Briefs should be self-contained. The agent receiving this should need zero follow-up questions.91- Default to over-specifying escalation conditions. Autonomous agents should fail loudly, not silently.92- Never include credentials in the brief. Reference environment variables or tool names, not actual keys/passwords.93- Keep it under 500 words. If the brief is longer, the task probably needs decomposition.9495## Source9697Nate's Newsletter, 2026-03-28 — "The open loop audit prompt that separates real delegation from simulated work"98Pattern: Structured handoff briefs for real delegation to autonomous agents.