Stats
Actions
Tags
From research-loop
Use when the user asks to initialize a research loop, create .research state, or start tracking a research idea from an idea file
How this skill is triggered — by the user, by Claude, or both
Slash command
/research-loop:research-loop-initThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Initialize a `.research/` directory from an idea file. Run this only once at the start of a research thread.
Initialize a .research/ directory from an idea file. Run this only once at the start of a research thread.
.research/ does not already exist. If it exists, stop and ask whether to resume instead.main, master, develop, or another non-experiment branch, suggest creating a branch but do not create one unless the user explicitly provides a branch name..research/
├── IDEA.md
├── tree.md
├── DASHBOARD.md
└── experiments/
IDEA.md:
# Research IDEA
## 动机
[1-2 paragraphs]
## 核心问题
- [question]
## 成功判据
[criteria]
## 初始假设
- H1: [summary]
## 参考资料
[links or notes from idea file]
tree.md using the hypothesis-tree format:
# Hypothesis Tree
## H1: [full hypothesis]
Status: 待验
Evidence: (empty)
Children: H1.1
### H1.1: [child hypothesis]
Status: 待验
Evidence: (empty)
Parent: H1
DASHBOARD.md:
# Research Dashboard
**IDEA**: [one-line idea]
**Active**: [count] hypotheses | **Last**: [YYYY-MM-DD]
## Active Hypotheses
- H1: [description] (待验)
## Next Steps
1. [next discriminating experiment or analysis]
待验, 进行中, 被支持, 被推翻 for hypothesis status.Status: in tree.md is plain text, never bold.(empty).Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub hamondyan/research-loop