用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/edwinhu/workflows --skill dev-design命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
Use for feature development and engineering changes.
This skill should be used when the user asks to 'start data analysis', 'plan a data project', 'explore this dataset', 'what should I analyze', 'set up a new study', or needs the data-science workflow.
Use when the user asks to 'run a work workflow', 'do this properly', 'clarify, plan, and verify this', 'small structured task', or 'don't just wing it' for a bounded task too small for a specialized workflow.
正在显示 SKILL.md
基于 SOC 职业分类
| name | dev-design |
| description | Internal architecture decision and native dev-plan authoring. |
| user-invocable | false |
| disable-model-invocation | true |
Announce: "I'm using dev-design to present architecture choices before creating the native plan."
!bun ${CLAUDE_SKILL_DIR}/../../scripts/load-constraints.ts dev-design
Use the conversational clarification and reconnaissance results; do not read or revive historical
.planning/PLAN.md, SPEC.md, reviews, progress files, or handoffs.
Present 2–3 feasible approaches, including a recommendation. For each, name the affected boundaries/files, compatibility and testing implications, implementation cost, and trade-offs. Obtain the user's explicit choice before entering Plan mode. A sole viable option still needs its trade-off explained and explicit acceptance.
Enter native Plan mode and generate exactly one executable Markdown plan. ExitPlanMode returns
its path; save that exact path and its returned identity for review and implementation. Do not
write a fixed .planning/PLAN.md, copy plan bytes, or create a visible approval marker.
The generated plan must contain these exact sections:
## Intent and Scope
## Requirements
## Chosen Architecture
## Testing Strategy and Real-Test Contract
## Implementation Tasks
## Requirement → Test Map
## Evidence Plan
## Review Surfaces
## Requirements assigns every in-scope behavioral requirement a stable REQ-NN identifier;
it states exclusions separately.## Chosen Architecture records the selected option, rationale, boundaries, interfaces, and
accepted trade-offs.## Testing Strategy and Real-Test Contract records the exact framework and commands, user
workflow, production protocol/transport, observable outcome, first failing test and expected
RED, plus the full-suite/E2E contract where applicable.## Implementation Tasks has one stable TASK-NN block per execution task. Each block must
include Dependencies, Work, Criteria, Outputs, Writable paths,
First failing test / RED expectation, Verify command, Instruction files, Model,
and Effort. Outputs and Writable paths are nonempty concrete repo-relative paths;
instruction files are absolute paths. Dependencies name TASK-NN IDs and form an acyclic DAG.
The work, criteria, outputs, writable paths, instruction files, model, and effort compile
directly to the shared TaskContract fields. Work requires RED before implementation and
GREEN through the exact verify command.## Requirement → Test Map maps every REQ-NN to one or more TASK-NN, named tests, and
evidence commands. No requirement can be prose-only or untested.## Evidence Plan defines command/runtime artifacts for every requirement and task.## Review Surfaces names security, correctness, regression, API/UI, and other independently
reviewable surfaces relevant to the change.Iron law: no implementation without a native approved plan. The generated plan bytes are
immutable after ExitPlanMode: changes to requirements, architecture, dependencies, real-test
contract, or evidence require a newly generated plan and receipt rollover.
After native approval, load skills/dev-plan-reviewer/SKILL.md; dispatch review with the exact
path returned by Plan mode, never a globbed or inferred plan.