research-knowledge

Name: research-knowledge
Author: shobcoder

578stars0forks

How to install research-knowledge?

shobcoder/shob/research-knowledge

$npx -y skills add shobcoder/shob --skill research-knowledge

Installs into the current project.

Summary

GOD MODE deep research skill. Multi-phase autonomous research loop: query decomposition → multi-source crawling → claim cross-referencing → conflict resolution → structured synthesis with inline citations. Use for any question that demands depth, sourcing, and verifiable accuracy

Files · 1

View on GitHub

SKILL.md

1	# Deep Research — GOD MODE
2
3	## Trigger
4	`/deep-research <query>` or when user says "research this deeply", "go deep on",
5	"full research report on", "investigate this thoroughly".
6
7	## Core Philosophy
8
9	> Raw search results are noise. Verified synthesis is signal.
10	> Every claim needs a source. Every conflict needs a resolution.
11	> A great deep research report is a structured intelligence brief, not a search summary.
12
13	## Architecture
14
15	```
16	Query
17	└── Phase 1: Decompose → Sub-questions
18	└── Phase 2: Parallel Search → Raw Sources
19	└── Phase 3: Crawl & Extract → Claims
20	└── Phase 4: Cross-Reference → Verify / Conflict
21	└── Phase 5: Synthesize → Report
22	└── Phase 6: Quality Gates → Deliver
23	```
24
25	---
26
27	## Phase 1 — Query Decomposition
28
29	Break the user's query into 3–7 atomic sub-questions. Each must be:
30	- Independently searchable
31	- Non-overlapping with others
32	- Ordered from foundational to advanced
33
34	Example:
35
36	> Query: "Is Company X profitable?"
37
38	Sub-questions:
39	1. What is Company X's current revenue model?
40	2. What are its reported ARR and revenue figures?
41	3. What is its burn rate and cost structure?
42	4. What do investors say about its path to profitability?
43	5. How does it compare to competitors on unit economics?
44
45	---
46
47	## Phase 2 — Multi-Source Search Strategy
48
49	For each sub-question, issue 2–4 targeted searches using varied query angles:
50
51	```
52	[primary term] [year]
53	[primary term] site:official OR filetype:pdf
54	[primary term] analysis OR breakdown OR report
55	[primary term] vs [competitor]
56	```
57
58	Source Priority Tiers:
59
60	\| Tier \| Type \| Trust Weight \|
61	\|------\|------\|-------------\|
62	\| 1 \| Official docs, SEC filings, company blogs, government data, peer-reviewed papers \| 1.0 \|
63	\| 2 \| Major news outlets (Reuters, Bloomberg, FT), industry analysts (Gartner, CB Insights) \| 0.85 \|
64	\| 3 \| Tech blogs, newsletters, podcasts \| 0.65 \|
65	\| 4 \| Forums, Reddit, social media \| 0.40 \|
66
67	Minimum sources per report: 8 unique domains
68	Target for complex topics: 15–25 sources
69
70	---
71
72	## Phase 3 — Deep Crawl and Extraction
73
74	For each source, fetch the full page (not just the snippet), then extract structured claims:
75
76	- Numerical facts (stats, dates, prices, percentages)
77	- Named entities (people, companies, products)
78	- Causal claims ("X caused Y because Z")
79	- Comparative claims ("A is better than B")
80
81	Tag each claim with source URL, publish date, tier rating, and a paraphrase or verbatim quote under 15 words.
82
83	Extraction template per source:
84
85	```yaml
86	source: https://example.com/article
87	published: 2026-04-12
88	tier: 2
89	claims:
90	- text: "Company reached $100M ARR in Q1 2026"
91	type: numerical
92	confidence: high
93	- text: "CEO stated profitability target by 2027"
94	type: causal
95	confidence: medium
96	```
97
98	---
99
100	## Phase 4 — Cross-Reference and Conflict Resolution
101
102	### 4a. Claim Clustering
103	Group identical or related claims across sources. If 3+ Tier 1–2 sources agree, mark as Verified.
104
105	### 4b. Conflict Detection
106	Flag claims where sources contradict each other:
107
108	```
109	CONFLICT DETECTED
110	Claim A: "Revenue is $50M ARR" — Source A, 2026-01
111	Claim B: "Revenue is $80M ARR" — Source B, 2026-03
112	Resolution: Use most recent Tier 1–2 source. Note discrepancy in report.
113	```
114
115	### 4c. Gap Detection
116	If a sub-question has zero Tier 1–2 sources, mark it `[unverified]` and flag it in the report.
117
118	### 4d. Confidence Scoring
119
120	```
121	Confidence = (sum of tier_weights x recency_factor) / num_claims
122
123	recency_factor:
124	< 30 days: 1.0
125	30–90 days: 0.9
126	3–12 months: 0.75
127	> 1 year: 0.60
128	```
129
130	---
131
132	## Phase 5 — Report Synthesis
133
134	### Report Structure
135
136	```markdown
137	# [Topic] — Deep Research Report
138
139	> Researched: [date] \| Sources: [N] \| Confidence: [X]% \| Sub-questions: [N]
140
141	---
142
143	## Executive Summary
144
145	2–4 sentence synthesis of the most important findings.
146	Lead with the single most important fact.
147
148	---
149
150	## Table of Contents
151
152	1. [Sub-question 1 title](#anchor)
153	2. [Sub-question 2 title](#anchor)
154	...
155	N. Conflicts and Uncertainties
156	N+1. Sources
157
158	---
159
160	## 1. [Sub-question Title]
161
162	### Finding
163	One clear, direct answer to the sub-question.
164
165	### Evidence
166	- [Claim] — Source: [Name], [Date], Tier 1
167	- [Claim] — Source: [Name], [Date], Tier 2
168	- [Claim] — Source: [Name], [Date], unverified
169
170	### Confidence: [X]% \| Coverage: [N] sources
171
172	---
173
174	## [Repeat for each sub-question]
175
176	---
177
178	## Conflicts and Uncertainties
179
180	\| Topic \| Claim A \| Claim B \| Resolution \|
181	\|-------\|---------\|---------\|------------\|
182	\| Revenue \| $50M (Source A) \| $80M (Source B) \| Use Source B (more recent) \|
183
184	---
185
186	## Knowledge Gaps
187
188	- [Field X]: No Tier 1–2 source

People who liked this also viewed

6 picks

Type

research-knowledge

578stars0forks

View on GitHub

How to install research-knowledge?

shobcoder/shob/research-knowledge

$npx -y skills add shobcoder/shob --skill research-knowledge

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Deep Research — GOD MODE
2
3	## Trigger
4	`/deep-research <query>` or when user says "research this deeply", "go deep on",
5	"full research report on", "investigate this thoroughly".
6
7	## Core Philosophy
8
9	> Raw search results are noise. Verified synthesis is signal.
10	> Every claim needs a source. Every conflict needs a resolution.
11	> A great deep research report is a structured intelligence brief, not a search summary.
12
13	## Architecture
14
15	```
16	Query
17	└── Phase 1: Decompose → Sub-questions
18	└── Phase 2: Parallel Search → Raw Sources
19	└── Phase 3: Crawl & Extract → Claims
20	└── Phase 4: Cross-Reference → Verify / Conflict
21	└── Phase 5: Synthesize → Report
22	└── Phase 6: Quality Gates → Deliver
23	```
24
25	---
26
27	## Phase 1 — Query Decomposition
28
29	Break the user's query into 3–7 atomic sub-questions. Each must be:
30	- Independently searchable
31	- Non-overlapping with others
32	- Ordered from foundational to advanced
33
34	Example:
35
36	> Query: "Is Company X profitable?"
37
38	Sub-questions:
39	1. What is Company X's current revenue model?
40	2. What are its reported ARR and revenue figures?
41	3. What is its burn rate and cost structure?
42	4. What do investors say about its path to profitability?
43	5. How does it compare to competitors on unit economics?
44
45	---
46
47	## Phase 2 — Multi-Source Search Strategy
48
49	For each sub-question, issue 2–4 targeted searches using varied query angles:
50
51	```
52	[primary term] [year]
53	[primary term] site:official OR filetype:pdf
54	[primary term] analysis OR breakdown OR report
55	[primary term] vs [competitor]
56	```
57
58	Source Priority Tiers:
59
60	\| Tier \| Type \| Trust Weight \|
61	\|------\|------\|-------------\|
62	\| 1 \| Official docs, SEC filings, company blogs, government data, peer-reviewed papers \| 1.0 \|
63	\| 2 \| Major news outlets (Reuters, Bloomberg, FT), industry analysts (Gartner, CB Insights) \| 0.85 \|
64	\| 3 \| Tech blogs, newsletters, podcasts \| 0.65 \|
65	\| 4 \| Forums, Reddit, social media \| 0.40 \|
66
67	Minimum sources per report: 8 unique domains
68	Target for complex topics: 15–25 sources
69
70	---
71
72	## Phase 3 — Deep Crawl and Extraction
73
74	For each source, fetch the full page (not just the snippet), then extract structured claims:
75
76	- Numerical facts (stats, dates, prices, percentages)
77	- Named entities (people, companies, products)
78	- Causal claims ("X caused Y because Z")
79	- Comparative claims ("A is better than B")
80
81	Tag each claim with source URL, publish date, tier rating, and a paraphrase or verbatim quote under 15 words.
82
83	Extraction template per source:
84
85	```yaml
86	source: https://example.com/article
87	published: 2026-04-12
88	tier: 2
89	claims:
90	- text: "Company reached $100M ARR in Q1 2026"
91	type: numerical
92	confidence: high
93	- text: "CEO stated profitability target by 2027"
94	type: causal
95	confidence: medium
96	```
97
98	---
99
100	## Phase 4 — Cross-Reference and Conflict Resolution
101
102	### 4a. Claim Clustering
103	Group identical or related claims across sources. If 3+ Tier 1–2 sources agree, mark as Verified.
104
105	### 4b. Conflict Detection
106	Flag claims where sources contradict each other:
107
108	```
109	CONFLICT DETECTED
110	Claim A: "Revenue is $50M ARR" — Source A, 2026-01
111	Claim B: "Revenue is $80M ARR" — Source B, 2026-03
112	Resolution: Use most recent Tier 1–2 source. Note discrepancy in report.
113	```
114
115	### 4c. Gap Detection
116	If a sub-question has zero Tier 1–2 sources, mark it `[unverified]` and flag it in the report.
117
118	### 4d. Confidence Scoring
119
120	```
121	Confidence = (sum of tier_weights x recency_factor) / num_claims
122
123	recency_factor:
124	< 30 days: 1.0
125	30–90 days: 0.9
126	3–12 months: 0.75
127	> 1 year: 0.60
128	```
129
130	---
131
132	## Phase 5 — Report Synthesis
133
134	### Report Structure
135
136	```markdown
137	# [Topic] — Deep Research Report
138
139	> Researched: [date] \| Sources: [N] \| Confidence: [X]% \| Sub-questions: [N]
140
141	---
142
143	## Executive Summary
144
145	2–4 sentence synthesis of the most important findings.
146	Lead with the single most important fact.
147
148	---
149
150	## Table of Contents
151
152	1. [Sub-question 1 title](#anchor)
153	2. [Sub-question 2 title](#anchor)
154	...
155	N. Conflicts and Uncertainties
156	N+1. Sources
157
158	---
159
160	## 1. [Sub-question Title]
161
162	### Finding
163	One clear, direct answer to the sub-question.
164
165	### Evidence
166	- [Claim] — Source: [Name], [Date], Tier 1
167	- [Claim] — Source: [Name], [Date], Tier 2
168	- [Claim] — Source: [Name], [Date], unverified
169
170	### Confidence: [X]% \| Coverage: [N] sources
171
172	---
173
174	## [Repeat for each sub-question]
175
176	---
177
178	## Conflicts and Uncertainties
179
180	\| Topic \| Claim A \| Claim B \| Resolution \|
181	\|-------\|---------\|---------\|------------\|
182	\| Revenue \| $50M (Source A) \| $80M (Source B) \| Use Source B (more recent) \|
183
184	---
185
186	## Knowledge Gaps
187
188	- [Field X]: No Tier 1–2 source

People who liked this also viewed

6 picks

Type