菜单
In-depth research on technical topics using web search, documentation, and codebase analysis. Use when asked to 'research', 'investigate', 'find out about', or explore unfamiliar technologies.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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/majiayu000/claude-skill-registry --skill deep-research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。
LLM token logprobs and calibration. Per-decision confidence, ECE, Brier, reliability diagrams, low-confidence triage.
Analyze LLM token logprobs and calibration. Use for per-decision confidence, ECE, Brier scores, reliability diagrams, and low-confidence triage.
回顾最近 N 天的 Claude Code 使用记录——扫描原始会话数据，按主题分组汇总"我都做了什么"，并从个人操作系统视角输出模式、风险与增删建议。当用户说 /recap、"看看我这几天做了什么"、"回顾一下我最近的会话"、"这两天我用 claude 干了啥"、"活动回顾" 时使用。
基于 SOC 职业分类
正在显示 SKILL.md
| name | deep-research |
| description | In-depth research on technical topics using web search, documentation, and codebase analysis. Use when asked to 'research', 'investigate', 'find out about', or explore unfamiliar technologies. |
| context | fork |
| allowed-tools | ["WebSearch","WebFetch","Read","Grep","Glob","Task"] |
Conduct comprehensive research on technical topics, synthesizing information from multiple sources. Since this runs in a forked context, explore extensively - only the final findings return to the main conversation.
Before diving in, understand:
| Priority | Source | Use For |
|---|---|---|
| 1 | Official docs | Authoritative information |
| 2 | GitHub repos | Real implementations, issues |
| 3 | Technical blogs | Best practices, gotchas |
| 4 | Stack Overflow | Common problems, solutions |
| 5 | Codebase analysis | How it relates to our code |
For Technology Evaluation:
1. Official documentation overview
2. GitHub - stars, activity, issues
3. Comparison articles (vs alternatives)
4. Real-world case studies
5. Our codebase - integration points
For Problem Solving:
1. Error message / symptom search
2. GitHub issues in relevant repos
3. Stack Overflow discussions
4. Official troubleshooting guides
5. Our codebase - similar patterns
For Best Practices:
1. Official style guides
2. Community conventions
3. Popular open-source examples
4. Our existing patterns
5. Team preferences (from CLAUDE.md)
Return structured findings:
# Research Report: [Topic]
**Research Question**: [what was investigated]
**Date**: [ISO date]
**Sources Consulted**: [count]
## Executive Summary
[3-5 sentences answering the research question]
## Key Findings
### Finding 1: [title]
**Confidence**: HIGH/MEDIUM/LOW
**Sources**: [list]
[detailed finding]
### Finding 2: [title]
...
## Recommendations
Based on research:
1. [recommendation with rationale]
2. [recommendation with rationale]
## Sources
- [Source 1](url) - [what it provided]
- [Source 2](url) - [what it provided]
## Open Questions
- [anything that couldn't be answered]
Before returning findings:
When researching, consider our stack: