harness-onboard

Automate agent session startup by reading claude-progress.txt and features.json.

1|Updated Feb 11, 2026
One-click install
npx skills add https://github.com/LauraFlorentin/skills-marketplace --skill harness-onboard
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: harness-onboard
Source: https://github.com/LauraFlorentin/skills-marketplace/tree/main/harness-engineer/skills/harness-onboard
Command: npx skills add https://github.com/LauraFlorentin/skills-marketplace --skill harness-onboard

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Automates the initial agent session setup in harness-enabled projects, eliminating the context gap when a new agent begins a run and ensuring consistent environment and starter guidance.

Core Features & Use Cases

  • Distinguishes initializer sessions from coding agent sessions to apply the appropriate prompt and scaffolding.
  • Validates the environment, reads recent history, and verifies prerequisites before starting work.
  • Reads claude-progress.txt and the feature list to determine the next task and primes the agent with the reasoning sandwich.
  • Reads features.json, marks the selected feature as in_progress, and prepares contextual artifacts (AGENTS.md, init.sh, layers.json, etc.) for the team's workflow.

Quick Start

Run the harness onboarding flow in a harness-enabled project to initialize the first agent session and prepare its context.

Frequently Asked Questions about harness-onboard

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

FAQPage Schema
How do I automate agent onboarding at session start in a harness-enabled project?

Agent onboarding is automated by reading claude-progress.txt and features.json, validating the environment, and seeding the agent with reasoning context to eliminate the initial context gap for new coding sessions.

What is the purpose of reading claude-progress.txt and features.json during session startup?

Reading claude-progress.txt and features.json determines the next task, marks the selected feature as in progress, and prepares contextual artifacts like AGENTS.md and layers.json for the team workflow.

Does the harness onboarding flow distinguish between initializer sessions and coding agent sessions?

Yes, the onboarding flow distinguishes initializer sessions from coding agent sessions to apply the appropriate prompt scaffolding and ensure correct environment validation before work begins.

Can I use harness-onboard to wire LocalContextMiddleware and prepare contextual artifacts?

Yes, harness-onboard wires LocalContextMiddleware to map structure and tools while preparing contextual artifacts including init.sh, AGENTS.md, and layers.json for the team's workflow.

How does the reasoning sandwich prime an AI coding agent during onboarding?

The reasoning sandwich primes the AI coding agent by seeding it with structured reasoning context derived from recent history and feature data, ensuring the environment and next feature are fully prepared.

What prerequisites do I need to run the harness onboarding flow for AI agents?

You need a harness-enabled project containing claude-progress.txt and features.json, along with prerequisites that the onboarding flow validates before seeding the agent and starting work.