modakbul-gongbang/harness
Developer toolsThis page was translated automatically. Technical details may differ from the original. View the original
Claude Code skills and practical materials for sessions. You can install only the necessary skills directly without plugins, and traditional plugin methods are also supported.
Overview
Claude Code skills and practical materials for sessions. You can install only the necessary skills directly without plugins, and traditional plugin methods are also supported. Harness Engineering is a technology for designing environments where AI agents can work well—hands-on practice and provides ready-to-use skills and resources. /scaffold selects language, framework, databases, and libraries according to project requirements and existing team environments, plans the structure and harness, and proceeds to actual implementation according to request. Technology is not determined solely by installed tools; the official compatibility basis for new dependencies is distinguished from the actual build results. In existing projects, the current structure and files in progress are preserved, and only the agreed scope is reinforced. For small projects, only the necessary choices are asked; if only planning is requested, implementation is not done. The implementation result is the selected DB· One example of the criteria using the API contract, a test, actual execution/verification commands, and necessary instructions/skills· Hooks. The verification report distinguishes between actual execution results and unverified items, and does not count the passing of the hook fixture as runtime activation success. If the necessary UI validation is blocked, it is not reported as completed. The procedure is scaffold SKILL.md, detailed verification criteria are verification.md, and the case for compliance checks is evals.json. This is a self-diagnosis reference material to explain harness design concepts in lectures. It is not the current /check-harness implementation criteria or result scorecard. The skill inspection procedure follows SKILL.md, and the criteria for each item are checklist.md. - - Can AI identify projects on its own - - CLAUDE.md, rules, incremental exposure - - Planning, delegation, orchestration - - Testing, review, quality control - - Learning, feedback loops "Harness Engineering – The Technology for Designing Environments Where AI Works Well" session presentation materials. 50 HTML slides + viewer.
Overview (machine translated)
Harness Engineering Claude Code skills and hands-on materials for sessions. You can install only the skills you need directly without plugins, and it also supports traditional plugin methods.
Harness Engineering is a technology for designing environments where AI agents can work well—hands-on practice and provides ready-to-use skills and resources.
Skills
| Skill | Description | Instructions |
|---|---|---|
| check-harness | Comprehensive Review of Development Environment, Context, Hook, and Verification System + Recommendation for Evidence-Based Improvements | /check-harness |
| scaffold | From technology and dependency selection to operational reference examples and harness configuration, planning and implementation | /scaffold |
| specify | Transforming Goals into Structured Implementation Plans (spec.md) | /specify "목표" |
| deep-interview | Socratic Requirements Interview (Based on Ambiguity Score) | /deep-interview "주제" |
Scaffold Usage Scope
/scaffold is about language, framework, DB, and library selection tailored to project requirements and the existing team environment, planning the structure and harness, and then proceeding with actual implementation according to requests.
Technology is not determined solely by installed tools; it distinguishes between the official compatibility rationale for new dependencies and the actual build results.
In existing projects, the current structure and files under work are preserved, and only the agreed scope is reinforced.
Small projects only ask for necessary choices, and if only plans are requested, they are not implemented.
The implementation results include one reference example for using the selected DB· API contract, tests, actual execution/verification commands, and necessary instructions, skills, Hooks. The verification report distinguishes between actual execution results and unverified items, and does not count hook fixture passage as runtime activation success. If the necessary UI validation is blocked, it is not reported as completed. The procedure is scaffold SKILL.md, the detailed verification criteria are verification.md, and the case of compliance inspection is evals.json.
Materials
Harness Engineering Checklist
materials/harness-checklist.md
This is a self-diagnosis reference material to explain harness design concepts in lectures.
This page was translated automatically. Technical details may differ from the original.
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add modakbul-gongbang/harness