# Harness Engineering Starter Kit

Use this kit to give an AI coding agent clear repository context, bounded work,
persistent state, and executable proof of completion.

## Start here

1. Copy `AGENTS.md` and `PROGRESS.md` to the root of your repository.
2. Replace every bracketed placeholder in `AGENTS.md`.
3. Adapt `feature-list.json` to your current work.
4. Complete `harness-audit.md` before adding more process.
5. Pick the relevant checks from `definition-of-done.md`.
6. Use `browser-verification.md` for user-facing flows.

## Included files

- `AGENTS.md`: a concise project map and working contract.
- `PROGRESS.md`: a session handoff that survives a new context window.
- `feature-list.json`: structured scope and acceptance criteria.
- `harness-audit.md`: a five-subsystem assessment.
- `definition-of-done.md`: an evidence-based completion checklist.
- `browser-verification.md`: a repeatable browser testing procedure.
- `repository-structures.md`: minimal and advanced repository layouts.
- `payment-flow-task.md`: a worked feature specification.

## Working rule

Do not adopt every file at once. Start with `AGENTS.md`, one bounded task, and
one verification command. Add the other files when a repeated failure shows
that they are needed.

This kit is provided by Agent Builders HQ and may be adapted for your projects.