agentleFS
YSAA1YSAA1/harness-workflow/skills/harness-builder/SKILL.md

harness-builder

Skill3 starsChanged 13 days agoMITView on GitHub

在目标项目搭或修 agent 工作台:项目入口、恢复面初始化、验证闸门、能力配置。触发:新项目第一次接入本套技能,或用户要搭/初始化/修工作台;日常任务不经过(任务级建档由 plan 自动做),已有工作台足够时直接回原任务。

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

Harness Builder

这是跨工作台问题的总控;交付与缺口相称的配置或建议,不把完整工作流当作所有任务前置条件。

流程

  1. 阅读与用户目标相关的入口、配置、脚本、恢复记录和实际错误。确认 canonical source、目标运行时及已有能力。
  2. 仅对真实缺口选方案,并识别可拆除的过期组件;简单修复可直接说明文件和检查,跨领域改造用一张 HARNESS RECOMMENDATION MATRIX 汇总缺口、证据、处理方式、负责人和验证。Required / Recommended / Deferred / Rejected 是建议等级,不是安装授权。
  3. helper 路由见下表。独立、边界明确的只读调查可并行;简单一行修改无需为每类文件完整启动另一套工作流,跨 helper 传递已有目标和授权。
  4. USER CHECKPOINT 仅用于实际缺少的授权或关键取舍。仅审计/推荐时只读;明确要求建立、修复或优化时完成范围内可逆编辑和必要验证,不要求用户批准每个精确 patch。新增外部操作或扩大配置范围需另行授权。
  5. 修改 canonical source;安装面受影响时同步 .claude-plugin/ manifest 与 README.md、docs/install.md,验证真实目标。不要默认往目标项目安装 hooks、MCP、子代理配置或所有模板。(本插件自带的 hooks/harness_guard.py 不写进目标项目:它随 Claude Code 插件加载,只在项目已有 .harness/index.md 时拦截写入 .harness/ 清单外的路径,停用插件即关闭。)目标项目选 harness 恢复面时,按 ../writing-for-agents/references/recovery_surface_policy.md 的合同:用 templates/index.md.j2、track.md.j2 建 .harness/,可用 lessons.md.j2 建带说明头的 lessons.md,AGENTS.md 的 Recovery 段取自 templates/AGENTS.md.j2;把本 skill 的 scripts/harness_lint.py 复制到目标项目 scripts/,并在检查入口(templates/check.sh.j2)调用它。生成的文本不引用目标项目不存在的东西;旧布局(work_index.md、state.md)交给 sweep 迁移。
  6. 相关检查通过且风险低时可以完成;复杂合同修改按 review 进一步审阅。若用户同时授权后续产品工作,继续该工作;仅工作台推荐不授权执行产品任务。

Helper Skill routing

缺口Helper Skill
持久指令与恢复面(AGENTS / CLAUDE / Cursor rules、.harness/ 索引与轨道、plan)writing-for-agents(Writing for Agents)
能力选型、寻找技能(hooks、MCP、子代理、插件只读推荐)find-skills

按需参考

  • 范围与授权:references/controller_discipline.md、references/install_policy.md。
  • 推荐矩阵:references/recommendation_matrix_policy.md;配置落点:references/decision_matrix.md。
  • 验证入口:references/verification_policy.md;架构守卫:references/architecture_enforcement_policy.md;漂移:references/anti_entropy.md。
  • 缺少 helper 时可用 references/capability_discovery_playbook.md;不要把 fallback 当重复必读。
  • 需要并行调查时:references/subagent_orchestration.md。

Recommended next skill

按尚未完成的实际工作选择 helper、implement、diagnose 或 review;只有本次记录需要整理时用 cleanup,否则结束。

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.