English | 한국어 | 日本語 | 简体中文

A prompt-first starter kit for turning repeated coding-agent mistakes into durable repository instructions, checks, memory, and evaluation.

Quick Start

Open the target repository with your coding agent and give it this prompt.

Show full adoption prompt

Use this kit to apply harness engineering to this repository:

https://github.com/baskduf/harness-starter-kit

Clone the kit into ./harness-starter-kit if it is not already present, read it,
then apply its prompt-first harness engineering workflow to this repository.

Requirements:
- Treat the current working directory as the target repository.
- Treat ./harness-starter-kit as read-only reference material after cloning.
- Inspect this repository before editing.
- Preserve existing architecture, tools, package manager, commands, docs, and
  conventions.
- Do not blindly copy templates.
- Add only the minimum useful harness pieces.
- Prefer updating existing docs/configs over duplicating them.
- Do not overwrite or delete existing files without explaining why.
- If I ask for /harness doctor, use
  ./harness-starter-kit/commands/harness-doctor.md.
- If I ask for /harness update after adoption, use
  ./harness-starter-kit/commands/harness-update.md to refresh the kit reference,
  record .harness/source.json, and selectively update target harness files
  without blindly overwriting existing files.
- If I ask for /harness refresh after adoption, use
  ./harness-starter-kit/commands/harness-refresh.md to review existing harness
  docs, rules, knowledge records, and checks for stale or duplicated guidance.
  Do not delete, archive, move, or rename files without my explicit approval for
  the specific files.
- If I ask for /harness review sub-agent, use
  ./harness-starter-kit/commands/harness-review.md and treat the request as
  explicit permission to use a read-only reviewer subagent when available and
  permitted by the active runtime and tool instructions. If unavailable,
  blocked, not permitted, or failed, report the fallback reason.
- If I ask for /harness review, use
  ./harness-starter-kit/commands/harness-review.md to review the current change
  set from an opposing harness-engineering perspective. Report findings,
  missing checks, overreach, durable memory gaps, and follow-up recommendations
  without modifying files unless I explicitly ask you to apply fixes after the
  review.

Expected result:
- project-specific AGENTS.md or updated existing agent instructions
- knowledge store if no equivalent exists
- lightweight drift checks based on this repo's real rules
- local verification commands using existing tools
- adoption report with files changed, checks to run, assumptions, remaining
  manual steps, failure memory, effectiveness measurement plan,
  normal/focused/manual gate placement, and whether
  ./harness-starter-kit should be removed, ignored, or kept before commit

For the full prompt and workflow details, see docs/prompts/apply-to-target-repo.md and docs/adoption-workflow.md.

💫 If this kit helps you, a GitHub star would be appreciated. 💫

Harness Theory

Harness engineering treats the repository as the durable operating environment for coding agents:

Harness = Instructions + Constraints + Feedback + Memory + Evaluation + Governance

Harness health is different from agent effectiveness. Harness Doctor can scan for durable repository evidence, but it cannot prove that agents make fewer mistakes. Measure that separately with task outcomes and effectiveness reports. See docs/theory/harness-engineering.md for the model.

Every recurring agent failure should be converted into at least one durable artifact: a clearer instruction, an automated constraint, a test or CI check, a decision or failure record, or a drift check.

Commands

The /harness ... names below are prompt conventions by default, not built-in editor commands. Type or paste them into your coding agent chat. In editors such as Cursor, they will not appear in the command palette unless you separately add matching custom slash commands.

Command	Use when
/harness doctor	Score baseline harness evidence without modifying files.
/harness update	Refresh the local ./harness-starter-kit reference after adoption.
/harness refresh	Review stale, duplicated, obsolete, or unused target harness guidance.
/harness review	Challenge the current change set before finishing.
/harness review sub-agent	Explicitly request a read-only reviewer subagent when the runtime permits it.

See commands/ for full workflows: doctor, update, refresh, and review.

How Adoption Works

Show adoption details

This is not primarily an automatic installer. The agent should inspect the target repository first, then adapt the smallest useful set of harness artifacts: instructions, enforceable constraints, feedback loops, durable memory, drift checks, and an adoption report. Follow docs/adoption-workflow.md and the prompt in docs/prompts/apply-to-target-repo.md.

Use the optional installer only when you want a skeleton before agent-driven adaptation. It copies profile snippets into docs/harness/profiles/<profile> for review; prompt-first adoption reads profiles from the cloned kit at harness-starter-kit/templates/profiles/<profile>.

python harness-starter-kit/scripts/apply_harness.py --target . --profile generic --dry-run

Profiles shown in the badges above are conservative reference snippets, not automatic migrations. See docs/profiles.md and docs/checklists/profile-absorption.md.

For the detailed documentation index, see docs/component-map.md. Common adoption references: docs/checklists/external-api-work.md, docs/checklists/decision-failure-memory.md, and docs/checklists/verification-scripts.md.

Validation coverage and local checks live in docs/validation.md. Lifecycle pilot details live in docs/examples/lifecycle-pilot-results.md. They do not prove that harness adoption reduces repeated agent mistakes. Use docs/evaluation.md, docs/templates/effectiveness-report.md, and docs/templates/task-outcome.yaml to measure comparable tasks, wrong-file edits, first-pass verification, and human rework.

Dogfood reports include TodayBus for a Next.js public-data target and Harness ERP for a Spring/Maven backend target. Both are harnessed-only benchmarks, not proof of effectiveness improvement.

Contributors

Thanks to everyone who has helped shape this kit through code, docs, reviews, examples, translations, and dogfooding.

License

This project is licensed under the MIT License.