Engineering
handoff-pack avatar

handoff-pack

Generates structured Handoff Pack prompts for delegating scoped coding tasks to Gemini with clear instructions, acceptance criteria, and output requirements.

Introduction

The Handoff Pack generator is a specialized skill for developer workflows, designed to create high-fidelity, structured prompts for delegating specific coding tasks from a primary agent (like Claude Code) to secondary models like Gemini. It acts as a bridge to ensure delegated work maintains the project's architectural standards, passes predefined quality gates, and adheres to strict formatting requirements. This tool is essential for teams managing complex codebases where task decomposition and cross-model collaboration are required. It helps prevent ambiguity during handoffs by mandating a clear session context, file-level scope definitions, and verifiable success criteria.

  • Automatically generates session context including Delegation IDs, project names, and branch information.

  • Enforces strict file-level scope constraints, preventing unauthorized modifications outside of the defined work area.

  • Uses templated task instructions that combine natural language objectives with code pattern examples for consistent implementation.

  • Integrates built-in acceptance criteria, including automated shell commands for TypeScript compilation (tsc), linting, and test suite execution.

  • Provides a robust guardrail system to identify when tasks are too complex, security-sensitive, or architecturally ambiguous, forcing an escalation instead of an incorrect guess.

  • Supports fix-up workflows for failed verifications, allowing for iterative, traceable corrections when delegated tasks fail initial validation.

  • Target audience: Developers using AI-assisted coding tools, DevOps engineers managing agentic workflows, and team leads overseeing automated task delegation.

  • Use cases: Automating code documentation tasks, performing routine refactoring within isolated modules, and offloading repetitive unit test generation to specialized models.

  • Required inputs: Task description, scope (file patterns), triage score, and relevant project commands (npm, pnpm, yarn).

  • Output expectations: A markdown-formatted Handoff Pack or Fix-up Pack containing clear instructions, bash commands for verification, and an exact response schema for completion reports.

Repository Stats

Stars
6
Forks
3
Open Issues
0
Language
Shell
Default Branch
main
Sync Status
Idle
Last Synced
May 3, 2026, 07:42 PM
View on GitHub