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/Yuvasee/samocode --skill task-definition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。
正在显示 SKILL.md
| name | task-definition |
| description | Interactive task definition with Q&A and documentation. |
Turn a raw task into a documented spec: grill the user until you reach a shared understanding, then record the decisions.
Session path: [SESSION_PATH from working memory] Task: $ARGUMENTS
Map the task as a design tree: every decision branches into the decisions that hang off it. Work the tree in rounds.
The frontier is every decision whose prerequisites are already settled — the questions you can ask now without guessing at answers you haven't heard yet. Ask the whole frontier in one round. A question whose answer depends on another question still open in this round belongs to a later round, not this one.
Format each question:
❓ **Q1** — **<question title>**: <question body, may span paragraphs and include options>
➡️ <your recommended answer + short justification>
The Q&A is done when the frontier is empty: every branch of the design tree visited, nothing left silently assumed. Say: "No more questions, we can move on!" and wait for the user to confirm shared understanding before documenting.
Create file: [SESSION_PATH]/[TIMESTAMP_FILE]-task-[task-slug].md
# Task: [title]
Date: [TIMESTAMP_LOG]
## Description
[What needs to be accomplished]
## Requirements
[Bullet list]
## Clarifications
### [Topic]
**Q:** [question]
**A:** [answer/decision]
**Rationale:** [why]
## Edge Cases
[Things to watch for]
## Success Criteria
[How we know it's done]
[SESSION_PATH]/_overview.md:
- [TIMESTAMP_ITERATION] Task defined: [title] -> [filename].md- [filename].md - Task: [title]cd [SESSION_DIR] && git add . && git commit -m "Task: [title]"/create-plan, /do, /dop2
Execute implementation tasks with different approaches (single, dual-agent, plan-based).
Code quality analysis through cleanup analysis, multi-perspective reviews, and explicit finding decisions.
Extract durable lessons from the current conversation, classify them, and route each to the narrowest enforceable code, test, skill, or scoped instruction source without growing a generic CLAUDE.md memory dump.