Skip to content
Back to skills

Handoff

ASecurity

Hand off in-progress work, an on-call shift, or an open incident to another person or agent using a fixed four-part structure, so nothing load-bearing is lost between one owner and the next. Use when work needs to change hands mid-flight, an on-call rotation is ending, or an open incident is passing to a new owner. Not for compacting a conversation so a fresh session can resume, which is a session-continuity concern.

  • 18 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added August 31, 2026
developmentgo

Works with

  • cli

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned September 22, 2026

npx -y skills add aethrox/doctrine --skill handoff --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Handoff?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Handoff
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/aethrox-handoff/badge)](https://www.skillsdirectory.com/skills/aethrox-handoff)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: handoff
description: Hand off in-progress work, an on-call shift, or an open incident to another person or agent using a fixed four-part structure, so nothing load-bearing is lost between one owner and the next. Use when work needs to change hands mid-flight, an on-call rotation is ending, or an open incident is passing to a new owner. Not for compacting a conversation so a fresh session can resume, which is a session-continuity concern.
---

# Handoff

A handoff fails when the receiver has to reconstruct context the sender already had. This skill's defining constraint, from SBAR (Situation-Background-Assessment-Recommendation); the structured handoff standard behind the Joint Commission's National Patient Safety Goal for handoff communication, adapted here from clinical and on-call practice: every handoff has exactly four parts, in that order, and nothing that belongs in one part gets buried inside another.

This covers handing off live, in-progress work between owners: an on-call shift, an open incident, or a task mid-flight. It is not the same job as compacting a conversation for a fresh agent session to resume (that's a session-continuity concern) or as writing a postmortem (that's `incident-response`'s closed-incident artifact, written after the fact rather than mid-handoff).

## Phase 1: Situation

State the current state in one or two lines: what's happening right now, headline only. A reader who stops after this line should still know whether this is urgent. Don't lead with history or analysis; the situation is the "what," not the "how we got here."

## Phase 2: Background

Give only the context that changes what the receiver does next: what's already been tried, what's already been ruled out, what decisions are already made and shouldn't be re-litigated. A background section that restates everything from the start forces the receiver to re-derive what actually matters; the discipline here is inclusion by relevance, not by completeness. If something was tried and failed, say so explicitly; the single most expensive handoff failure is the receiver re-attempting a dead end because nobody mentioned it was already dead.

## Phase 3: Assessment

State the current read of the problem, not just facts, but the sender's judgment: how serious is this, what's the actual root concern, what's still genuinely unknown versus just not yet written down. This is the part a bare status report skips, and it's the part that carries the most value, because it's the sender's accumulated judgment, not just a log of events.

## Phase 4: Recommendation

Give a concrete next action, not just a description of the problem. If there's a decision only the receiver can make, name it explicitly rather than leaving it implicit in the background. A handoff that ends at Assessment leaves the receiver to invent their own recommendation from scratch, often duplicating judgment the sender already reached.

## Phase 5: Confirm receipt

A handoff isn't complete when it's sent. Closed-loop communication (the same principle SBAR is built on) means the receiver reads back what they understood before the sender disengages. For a live on-call or incident handoff, get an explicit acknowledgment of the Assessment and Recommendation specifically, not just "got it"; a receiver who can restate the recommendation in their own words has actually received it.

## Done when

- [ ] Situation is a short headline, not a summary of everything that follows.
- [ ] Background includes only what changes the receiver's next action, and explicitly names anything already tried and ruled out.
- [ ] Assessment states the sender's judgment, not just a list of facts.
- [ ] Recommendation names a concrete next action and any decision only the receiver can make.
- [ ] Receipt was confirmed: the receiver restated the Assessment and Recommendation, not just acknowledged receiving a message.
- [ ] Generated prose uses commas, colons, periods, or reworded sentences instead of em dashes.

Files in this skill

  • SKILL.md3.8 KB
  • agents/openai.yaml116 B

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.