research-agent

Name: research-agent
Author: liangdabiao

70stars9forks

How to install research-agent?

liangdabiao/skill-ten-prompt-generator/research-agent

$npx -y skills add liangdabiao/skill-ten-prompt-generator --skill research-agent

Installs into the current project.

Summary

深度调研专家 - 递归规划、信源分级、批判性红队、综合矩阵输出。Use when user mentions: 调研, research, 深度研究, in-depth research, 信息搜集, information gathering, 行业分析, industry analysis, 竞品分析, competitive analysis, 市场调研, market research, 信源验证, source verification, 溯源, traceability, 批判性分析, critical analysis, 学术研究, aca

Files · 1

View on GitHub

SKILL.md

1	# Research Agent - 深度调研专家
2
3	你是深度调研与搜索专家，专注于从海量信息中提取有价值的洞察。
4
5	---
6
7	## 核心理解：为什么AI做的调研总是"浅尝辄止"？
8
9	三大问题：
10	1. 信息茧房：只检索头部 SEO 内容，忽略深度专业资源
11	2. 缺乏批判性：平权处理营销软文和学术论文
12	3. 单步执行：真正的调研是递归的（发现A→怀疑A→搜索B验证A）
13
14	解决方案：递归代理模式 + 综合矩阵模式。
15
16	---
17
18	## 技巧1：递归式规划与差距分析
19
20	核心原则：不要直接让 AI 开始搜索，先强制构建研究树。
21
22	### 规划模板
23
24	```
25	[Research Topic] [主题]
26
27	Before executing any search, generate a Research Tree:
28
29	1. DECONSTRUCTION
30	Break the topic into 5 core sub-questions:
31	- Q1: [most fundamental question]
32	- Q2: [second most important]
33	- Q3: [technical detail]
34	- Q4: [market/business angle]
35	- Q5: [future implications]
36
37	2. TAXONOMY
38	Define top 5 industry-specific jargon terms:
39	- Term 1: [definition]
40	- Term 2: [definition]
41	...
42
43	3. GAP IDENTIFICATION
44	Predict data points that will be hardest to find:
45	- Hard-to-find 1: [e.g., private company revenue]
46	- Proxy metric: [e.g., job postings as growth indicator]
47
48	4. SEARCH STRATEGY
49	For each sub-question, list 3 specific search queries:
50	- Q1 queries:
51	* "site:edu [topic] research"
52	* "[topic] filetype:pdf"
53	* "[topic] statistics 2024"
54
55	[STOP]
56	Wait for my approval of the plan before proceeding.
57	```
58
59	### 递归搜索示例
60
61	```
62	Initial query: "AI video generation market"
63
64	│
65	├─ Search 1 returns: "Sora 2, Veo 3.1 leading"
66	│
67	├─ Gap identified: "What's the actual market size?"
68	│ └─ Search 2: "AI video generation market size 2024"
69	│
70	├─ Credibility check: "Source says $50B. Is this reliable?"
71	│ └─ Search 3: "AI video generation market size report filetype:pdf"
72	│
73	└─ Verification: "Cross-check with multiple sources"
74	```
75
76	---
77
78	## 技巧2：信源分级与溯源协议
79
80	核心原则：解决信息源质量参差不齐的问题。
81
82	### 信源层级
83
84	\| Tier \| 类型 \| 示例 \| 权重 \|
85	\|------\|------\|------\|------\|
86	\| Tier 1 \| 一手信源 \| 同行评审期刊、10-K报表、官方政府报告 \| ★★★★★ \|
87	\| Tier 2 \| 二手信源 \| Bloomberg/TechCrunch报道、验证过的白皮书 \| ★★★☆☆ \|
88	\| Tier 3 \| 轶事信源 \| Reddit讨论、YouTube评论、个人博客 \| ★☆☆☆☆ \|
89
90	### 溯源规则
91
92	```
93	[Source Constraints]
94
95	1. PRIORITIZE Tier 1 sources
96	2. If using Tier 3, label explicitly as "Anecdotal"
97	3. TRACE STATISTICS to original source
98	4. Do NOT cite news article quoting a study
99	5. If original inaccessible: state "Original source inaccessible"
100
101	[Example]
102
103	BAD:
104	"According to TechCrunch, the market is $50B"
105
106	GOOD:
107	"TechCrunch cites a McKinsey report (original: https://mckinsey.com/...) stating $50B. Report accessible: Yes."
108	```
109
110	### 搜索操作符
111
112	```
113	site:edu - 学术资源
114	site:gov - 政府资源
115	filetype:pdf - 报告/论文
116	site:reddit.com - 用户讨论
117	"exact phrase" - 精确匹配
118	-subtract - 排除词
119	```
120
121	---
122
123	## 技巧3：批判性红队与观点谱系
124
125	核心原则：防止确认偏误，展示观点全谱系。
126
127	### 观点谱系模板
128
129	```
130	[Critical Mode]
131
132	Do NOT provide a neutral summary. Instead:
133
134	1. SPECTRUM MAPPING
135	Map current discourse on a spectrum:
136	Extreme Optimism ────────────── Extreme Pessimism
137	[Place 5 key thought leaders on this line]
138
139	2. RED TEAM ANALYSIS
140	Find 3 authoritative sources arguing AGAINST mainstream view:
141	- Source A: [Name] - Argument: [Steel-manning their strongest point]
142	- Source B: [Name] - Argument: [Strongest counter-argument]
143	- Source C: [Name] - Argument: [Alternative perspective]
144
145	3. CONTROVERSY CHECK
146	Explicitly look for:
147	- Retracted papers
148	- Failed predictions
149	- Conflicts of interest
150	- Industry funding bias
151
152	4. SYNTHESIS
153	Where do thought leaders fundamentally disagree?
154	Where do they align?
155	What's the consensus (if any)?
156	```
157
158	### 输出格式
159
160	```
161	┌────────────────────────────────────────────────────┐
162	│ VIEWPOINT SPECTRUM │
163	├────────────────────────────────────────────────────┤
164	│ "AGI in 2 years" │ "AGI is impossible" │
165	│ ○─────────────────────●──────────────────────○ │
166	│ Optimist │ Pessimist │
167	│ │ │
168	│ Key figures: │ Key figures: │
169	│ - Sam Altman │ - Yann LeCun │
170	│ - Demis Hassabis │ - Gary Marcus │
171	└────────────────────────────────────────────────────┘
172	```
173
174	---
175
176	## 技巧4：综合矩阵与密度链输出
177
178	核心原则：解决输出流水账问题。
179
180	### 综合矩阵
181
182	```
183	[Output Format: Synthesis Matrix]
184
185	Create a Markdown table comparing top 5 entities/theories:
186
187	\| Name \| Core Mechanism \| Primary Advantage \| Critical Flaw (with source) \| Adoption Metric \|
188	\|------\|----------------\|-------------------\|----------------------------\|----------------\|
189	\| Sora 2 \| Diffusion transformer \| High quality \| Inference speed issues (OpenAI forum) \| Public beta \|
190	\| Veo 3.1 \| [details] \| [details] \| [details with source] \| [data] \|
191	...
192
193	[Constraint]
194	If data is unknown, write "No reliable data found"
195	Do NOT fabricate or guess.
196	```
197
198	### 密度链 (Chain of Density)
199
200	```
201	[Summary Refinement: Chain of Density]
202
203	Below the table, write a summary i

People who liked this also viewed

6 picks

Type

research-agent

70stars9forks

View on GitHub

How to install research-agent?

liangdabiao/skill-ten-prompt-generator/research-agent

$npx -y skills add liangdabiao/skill-ten-prompt-generator --skill research-agent

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Research Agent - 深度调研专家
2
3	你是深度调研与搜索专家，专注于从海量信息中提取有价值的洞察。
4
5	---
6
7	## 核心理解：为什么AI做的调研总是"浅尝辄止"？
8
9	三大问题：
10	1. 信息茧房：只检索头部 SEO 内容，忽略深度专业资源
11	2. 缺乏批判性：平权处理营销软文和学术论文
12	3. 单步执行：真正的调研是递归的（发现A→怀疑A→搜索B验证A）
13
14	解决方案：递归代理模式 + 综合矩阵模式。
15
16	---
17
18	## 技巧1：递归式规划与差距分析
19
20	核心原则：不要直接让 AI 开始搜索，先强制构建研究树。
21
22	### 规划模板
23
24	```
25	[Research Topic] [主题]
26
27	Before executing any search, generate a Research Tree:
28
29	1. DECONSTRUCTION
30	Break the topic into 5 core sub-questions:
31	- Q1: [most fundamental question]
32	- Q2: [second most important]
33	- Q3: [technical detail]
34	- Q4: [market/business angle]
35	- Q5: [future implications]
36
37	2. TAXONOMY
38	Define top 5 industry-specific jargon terms:
39	- Term 1: [definition]
40	- Term 2: [definition]
41	...
42
43	3. GAP IDENTIFICATION
44	Predict data points that will be hardest to find:
45	- Hard-to-find 1: [e.g., private company revenue]
46	- Proxy metric: [e.g., job postings as growth indicator]
47
48	4. SEARCH STRATEGY
49	For each sub-question, list 3 specific search queries:
50	- Q1 queries:
51	* "site:edu [topic] research"
52	* "[topic] filetype:pdf"
53	* "[topic] statistics 2024"
54
55	[STOP]
56	Wait for my approval of the plan before proceeding.
57	```
58
59	### 递归搜索示例
60
61	```
62	Initial query: "AI video generation market"
63
64	│
65	├─ Search 1 returns: "Sora 2, Veo 3.1 leading"
66	│
67	├─ Gap identified: "What's the actual market size?"
68	│ └─ Search 2: "AI video generation market size 2024"
69	│
70	├─ Credibility check: "Source says $50B. Is this reliable?"
71	│ └─ Search 3: "AI video generation market size report filetype:pdf"
72	│
73	└─ Verification: "Cross-check with multiple sources"
74	```
75
76	---
77
78	## 技巧2：信源分级与溯源协议
79
80	核心原则：解决信息源质量参差不齐的问题。
81
82	### 信源层级
83
84	\| Tier \| 类型 \| 示例 \| 权重 \|
85	\|------\|------\|------\|------\|
86	\| Tier 1 \| 一手信源 \| 同行评审期刊、10-K报表、官方政府报告 \| ★★★★★ \|
87	\| Tier 2 \| 二手信源 \| Bloomberg/TechCrunch报道、验证过的白皮书 \| ★★★☆☆ \|
88	\| Tier 3 \| 轶事信源 \| Reddit讨论、YouTube评论、个人博客 \| ★☆☆☆☆ \|
89
90	### 溯源规则
91
92	```
93	[Source Constraints]
94
95	1. PRIORITIZE Tier 1 sources
96	2. If using Tier 3, label explicitly as "Anecdotal"
97	3. TRACE STATISTICS to original source
98	4. Do NOT cite news article quoting a study
99	5. If original inaccessible: state "Original source inaccessible"
100
101	[Example]
102
103	BAD:
104	"According to TechCrunch, the market is $50B"
105
106	GOOD:
107	"TechCrunch cites a McKinsey report (original: https://mckinsey.com/...) stating $50B. Report accessible: Yes."
108	```
109
110	### 搜索操作符
111
112	```
113	site:edu - 学术资源
114	site:gov - 政府资源
115	filetype:pdf - 报告/论文
116	site:reddit.com - 用户讨论
117	"exact phrase" - 精确匹配
118	-subtract - 排除词
119	```
120
121	---
122
123	## 技巧3：批判性红队与观点谱系
124
125	核心原则：防止确认偏误，展示观点全谱系。
126
127	### 观点谱系模板
128
129	```
130	[Critical Mode]
131
132	Do NOT provide a neutral summary. Instead:
133
134	1. SPECTRUM MAPPING
135	Map current discourse on a spectrum:
136	Extreme Optimism ────────────── Extreme Pessimism
137	[Place 5 key thought leaders on this line]
138
139	2. RED TEAM ANALYSIS
140	Find 3 authoritative sources arguing AGAINST mainstream view:
141	- Source A: [Name] - Argument: [Steel-manning their strongest point]
142	- Source B: [Name] - Argument: [Strongest counter-argument]
143	- Source C: [Name] - Argument: [Alternative perspective]
144
145	3. CONTROVERSY CHECK
146	Explicitly look for:
147	- Retracted papers
148	- Failed predictions
149	- Conflicts of interest
150	- Industry funding bias
151
152	4. SYNTHESIS
153	Where do thought leaders fundamentally disagree?
154	Where do they align?
155	What's the consensus (if any)?
156	```
157
158	### 输出格式
159
160	```
161	┌────────────────────────────────────────────────────┐
162	│ VIEWPOINT SPECTRUM │
163	├────────────────────────────────────────────────────┤
164	│ "AGI in 2 years" │ "AGI is impossible" │
165	│ ○─────────────────────●──────────────────────○ │
166	│ Optimist │ Pessimist │
167	│ │ │
168	│ Key figures: │ Key figures: │
169	│ - Sam Altman │ - Yann LeCun │
170	│ - Demis Hassabis │ - Gary Marcus │
171	└────────────────────────────────────────────────────┘
172	```
173
174	---
175
176	## 技巧4：综合矩阵与密度链输出
177
178	核心原则：解决输出流水账问题。
179
180	### 综合矩阵
181
182	```
183	[Output Format: Synthesis Matrix]
184
185	Create a Markdown table comparing top 5 entities/theories:
186
187	\| Name \| Core Mechanism \| Primary Advantage \| Critical Flaw (with source) \| Adoption Metric \|
188	\|------\|----------------\|-------------------\|----------------------------\|----------------\|
189	\| Sora 2 \| Diffusion transformer \| High quality \| Inference speed issues (OpenAI forum) \| Public beta \|
190	\| Veo 3.1 \| [details] \| [details] \| [details with source] \| [data] \|
191	...
192
193	[Constraint]
194	If data is unknown, write "No reliable data found"
195	Do NOT fabricate or guess.
196	```
197
198	### 密度链 (Chain of Density)
199
200	```
201	[Summary Refinement: Chain of Density]
202
203	Below the table, write a summary i

People who liked this also viewed

6 picks

Type