evanfang0054/agent-harness

brainstorming

You MUST use this before any creative work — features, components, behavior changes. Explores intent, requirements, and design before implementation.

finishing-a-development-branch

Use when implementation is done and tests pass, and the user must choose how to integrate or retain completed branch work.

receiving-code-review

Use when receiving code review feedback. Verify before implementing; technical correctness over blind agreement.

subagent-driven-development

Use when executing implementation plans with independent tasks in the current session

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes

test-driven-development

Use when implementing any feature or bugfix, before writing implementation code

using-agent-harness

Use when starting any conversation — establishes skill discovery and requires Skill tool invocation before ANY response.

writing-plans

Use when you have a spec or requirements for a multi-step task, before touching code

writing-skills

Use when creating new skills, editing existing skills, or verifying skills work before deployment

dispatching-parallel-agents

Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies

executing-plans

Use when you have a written implementation plan to execute iteratively with review checkpoints

requesting-code-review

Use when completing tasks, implementing major features, or before merging to verify work meets requirements

verification-before-completion

Use before claiming work is complete or creating PRs. Run verification commands first; evidence before assertions.

domain-modeling

Build and sharpen the project's domain model. Use when domain terminology needs defining or sharpening, an architectural decision crystallizes, or another skill needs to maintain the domain glossary during design work.

harness-init

Use when initializing agent-harness in a new project or reconfiguring for a specific tech stack.

computational-sensors

Use when running deterministic checks (lint/typecheck/test/coverage) before semantic review, or when verification needs computational evidence.

documentation-sync

Use after code changes, before PR or merge. Sync README/ARCHITECTURE/CHANGELOG/CONTRIBUTING with the diff. Trigger: 'sync docs'.

fix-issues-and-pr

Fix existing GitHub issues via SDD and ship as one PR. Trigger on "修

Source text: Mixed languages

gate-driven-test-design

Use after brainstorming produces an approved design spec, before writing-plans — derives a risk-based test coverage tree (Gates + Assertions).

Source text: Chinese

generate-issues

Analyze sessions and file GitHub issues, no code changes. Trigger on "分析会话提 issue", "只生成 issue 不修复", "盘点会话问题".

Source text: Chinese

harness-design

用HTML做高保真原型、Demo、幻灯片、动画、设计变体。触发词：做原型、设计Demo、交互原型、UI mockup、prototype、app原型、iOS/移动端原型、设计风格/方向、配色/视觉风格、做个HTML页面、导出MP4/GIF、动画Demo、设计评审。

Source text: Chinese

harness-optimizer

Use when用户提供会话 ID，想复盘会话或基于该会话优化指定的项目、workflow、skill 或 harness。

Source text: Chinese

loop-detection

Use when an agent is stuck editing the same file repeatedly without converging, or when verification requires doom loop analysis.

plan-ceo-review

CEO/founder-mode plan review — rethink the problem, challenge premises, find the 10-star product. Trigger on "think bigger", "expand scope", "strategy review", or "rethink this".

plan-eng-review

Eng manager-mode plan review — lock in architecture, data flow, edge cases, test coverage. Trigger on "review the architecture", "engineering review", or "lock in the plan".

post-deploy-monitoring

Verify production health after deploy. Monitor errors, performance, page failures. Trigger: 'monitor deploy', 'check production'.

qa-testing

Systematically QA test a web app and fix bugs. Trigger on "qa", "test this site", "find bugs", "fix what's broken", or "does this work?".

session-learnings

Capture and reuse project knowledge across sessions. Use when discovering patterns/pitfalls, after debugging, or when user corrects your approach.

sprint-contract

Use after brainstorming produces a spec and before writing-plans — negotiates explicit Definition of Done to prevent ambiguity.

trace-analysis

Use during retrospective or when analyzing recurring failure patterns across sessions via learnings data.

retrospective

Engineering retrospective for analyzing recent work. Use when asked to 'do a retro', 'review the week', 'what did we accomplish', or at the end of a sprint/milestone. Analyzes commits, learnings, and patterns to improve future work.

office-hours

YC Office Hours — decision layer skill that answers "should we build this?" Startup mode: six forcing questions exposing demand reality, status quo, desperate specificity, narrowest wedge, observation, and future-fit. Builder mode: quick decision framework…