multi-agent-handoff

Plan and document handoffs and lane contracts between specialized AI agents.

2|Updated Aug 15, 2026
One-click install
npx skills add https://github.com/Arenukvern/shortly --skill multi-agent-handoff-arenukvern
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: multi-agent-handoff
Source: https://github.com/Arenukvern/shortly/tree/main/.agents/skills/multi-agent-handoff
Command: npx skills add https://github.com/Arenukvern/shortly --skill multi-agent-handoff-arenukvern

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Multi-agent workflows lose context when work passes between foreman, worker, and reviewer agents, leading to unverified claims, duplicated effort, and silently dropped subagent results. ## Core Features & Use Cases - Handoff Document Template: Provides a structured HANDOFF.md format with goal, done, next steps, constraints, verification, and validation status sections. - Parallel Batch Contracts: Defines lane tables with write sets, forbidden paths, native gates, claim ceilings, and terminal states for dispatching parallel subagents. - Landing Phase Discipline: Requires temp-clone proofs to be landed as minimal source-owned diffs in the owner checkout with the native gate rerun before claims are strengthened. - Use Case: When splitting a large refactor across explorer, implementer, and reviewer subagents, use this Skill to write a baton document that preserves the original goal, bounds each lane's write authority, and closes every lane with an explicit terminal state. ## Quick Start Ask the agent to create a HANDOFF.md using the multi-agent handoff template before dispatching subagents for the current task.

Frequently Asked Questions about multi-agent-handoff

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I hand off work between AI agents without losing context?▼

Write a HANDOFF.md baton containing the goal, completed items, ordered next steps (max seven), constraints, verification commands, and validation status. The receiving agent executes only the Next section and updates Done before re-handing off.

How to coordinate parallel subagents on independent repo tasks?▼

Use a parallel batch contract with a lane table declaring each agent's scope, exact write set, forbidden paths, native gate, and terminal state. Keep scopes non-overlapping and declare write ownership before implementation begins.

Does a fix proven in a temp clone count as repo evidence?▼

No. Temp-clone proof is only accepted_as_input until the minimal source-owned diff is applied to the owner checkout and the native gate is rerun there. Record source_owner_status as temp_only until the owner gate passes.

What happens when a subagent times out or vanishes?▼

The parent synthesis must record the lane with an explicit terminal state such as timed_out, partial, blocked, or superseded, and downgrade the overall claim. Vanished agents must never be silently absorbed as integrated evidence.

When should I not use a multi-agent handoff?▼

Skip it for single-step local work like running one formatter after a small edit, where the repo's native workflow suffices. Handoffs longer than one screen should be split into references or issues.