deep-research

Name: deep-research
Author: proyecto26

20installs2forks

How to install deep-research?

proyecto26/sherlock-ai-plugin/deep-research

$npx -y skills add proyecto26/sherlock-ai-plugin --skill deep-research

Installs into the current project.

Summary

Generate format-controlled research reports with evidence tracking, citations, and iterative review. This skill should be used when users request a research report, literature review, market or industry analysis, competitive landscape, policy or technical brief, or require a stri

Files · 1

View on GitHub

SKILL.md

1	# Deep Research
2
3	Create high-fidelity research reports with strict format control, evidence mapping, and multi-pass synthesis.
4
5	## Quick Start
6
7	1. Clarify the report spec and format contract
8	2. Build a research plan and query set
9	3. Collect evidence with the deepresearch tool (multi-pass if needed)
10	4. Triage sources and build an evidence table
11	5. Draft the full report in multiple complete passes (parallel subagents)
12	6. UNION merge, enforce format compliance, verify citations
13	7. Present draft for human review and iterate
14
15	## Core Workflow
16
17	Copy this checklist and track progress:
18
19	```
20	Deep Research Progress:
21	- [ ] Step 1: Intake and format contract
22	- [ ] Step 2: Research plan and query set
23	- [ ] Step 3: Evidence collection (deepresearch tool)
24	- [ ] Step 4: Source triage and evidence table
25	- [ ] Step 5: Outline and section map
26	- [ ] Step 6: Multi-pass full drafting (parallel subagents)
27	- [ ] Step 7: UNION merge and format compliance
28	- [ ] Step 8: Evidence and citation verification
29	- [ ] Step 9: Present draft for human review and iterate
30	```
31
32	### Step 1: Intake and Format Contract
33
34	Establish the report requirements before any research:
35
36	- Confirm audience, purpose, scope, time range, and geography
37	- Lock output format: Markdown, DOCX, slides, or user-provided template
38	- Capture required sections and exact formatting rules
39	- Confirm citation style (footnotes, inline, numbered, APA, etc.)
40	- Confirm length targets per section
41	- Ask for any existing style guide or sample report
42
43	Create a concise report spec file:
44
45	```
46	Report Spec:
47	- Audience:
48	- Purpose:
49	- Scope:
50	- Time Range:
51	- Geography:
52	- Required Sections:
53	- Section Formatting Rules:
54	- Citation Style:
55	- Output Format:
56	- Length Targets:
57	- Tone:
58	- Must-Include Sources:
59	- Must-Exclude Topics:
60	```
61
62	If a user provides a template or an example report, treat it as a hard constraint and mirror the structure.
63
64	### Step 2: Research Plan and Query Set
65
66	Define the research strategy before calling tools:
67
68	- Break the main question into 3-7 subquestions
69	- Define key entities, keywords, and synonyms
70	- Identify primary sources vs secondary sources
71	- Define disqualifiers (outdated, low quality, opinion-only)
72	- Assemble a query set per section
73
74	Use [references/research_plan_checklist.md](references/research_plan_checklist.md) for guidance.
75
76	### Step 3: Evidence Collection (Deepresearch Tool)
77
78	Use the deepresearch tool to collect evidence and citations.
79
80	- Run multiple complete passes if coverage is uncertain
81	- Vary query phrasing to reduce blind spots
82	- Preserve raw tool output in files for traceability
83
84	File structure (recommended):
85	```
86	<output_dir>/research/<topic-name>/
87	deepresearch_pass1.md
88	deepresearch_pass2.md
89	deepresearch_pass3.md
90	```
91
92	If deepresearch is unavailable, rely on user-provided sources only and state limitations explicitly.
93
94	### Step 4: Source Triage and Evidence Table
95
96	Normalize and score sources before drafting:
97
98	- De-duplicate sources across passes
99	- Score sources using [references/source_quality_rubric.md](references/source_quality_rubric.md)
100	- Build an evidence table mapping claims to sources
101
102	Evidence table minimum columns:
103
104	- Source ID
105	- Title
106	- Publisher
107	- Date
108	- URL or reference
109	- Quality tier (A/B/C)
110	- Notes
111
112	### Step 5: Outline and Section Map
113
114	Create an outline that enforces the format contract:
115
116	- Use the template in [references/research_report_template.md](references/research_report_template.md)
117	- Produce a section map with required elements per section
118	- Confirm ordering and headings match the report spec
119
120	### Step 6: Multi-Pass Full Drafting (Parallel Subagents)
121
122	Avoid single-pass drafting; generate multiple complete reports, then merge.
123
124	#### Preferred Strategy: Parallel Subagents (Complete Draft Each)
125
126	Use the Task tool to spawn parallel subagents with isolated context. Each subagent must:
127
128	- Load the report spec, outline, and evidence table
129	- Draft the FULL report (all sections)
130	- Enforce formatting rules and citation style
131
132	Implementation pattern:
133	```
134	Task(subagent_type="general-purpose", prompt="Draft complete report ...", run_in_background=false) -> version1.md
135	Task(subagent_type="general-purpose", prompt="Draft complete report ...", run_in_background=false) -> version2.md
136	Task(subagent_type="general-purpose", prompt="Draft complete report ...", run_in_background=false) -> version3.md
137	```
138
139	Write drafts to files, not conversation context:
140	```
141	<output_dir>/intermediate/<topic-name>/version1.md
142	<output_dir>/intermediate/<topic-name>/version2.md
143	<output_dir>/intermediate/<topic-name>/version3.md
144	```
145
146	### Step

People who liked this also viewed

6 picks

Type

deep-research

20installs2forks

View on GitHub

How to install deep-research?

proyecto26/sherlock-ai-plugin/deep-research

$npx -y skills add proyecto26/sherlock-ai-plugin --skill deep-research

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Deep Research
2
3	Create high-fidelity research reports with strict format control, evidence mapping, and multi-pass synthesis.
4
5	## Quick Start
6
7	1. Clarify the report spec and format contract
8	2. Build a research plan and query set
9	3. Collect evidence with the deepresearch tool (multi-pass if needed)
10	4. Triage sources and build an evidence table
11	5. Draft the full report in multiple complete passes (parallel subagents)
12	6. UNION merge, enforce format compliance, verify citations
13	7. Present draft for human review and iterate
14
15	## Core Workflow
16
17	Copy this checklist and track progress:
18
19	```
20	Deep Research Progress:
21	- [ ] Step 1: Intake and format contract
22	- [ ] Step 2: Research plan and query set
23	- [ ] Step 3: Evidence collection (deepresearch tool)
24	- [ ] Step 4: Source triage and evidence table
25	- [ ] Step 5: Outline and section map
26	- [ ] Step 6: Multi-pass full drafting (parallel subagents)
27	- [ ] Step 7: UNION merge and format compliance
28	- [ ] Step 8: Evidence and citation verification
29	- [ ] Step 9: Present draft for human review and iterate
30	```
31
32	### Step 1: Intake and Format Contract
33
34	Establish the report requirements before any research:
35
36	- Confirm audience, purpose, scope, time range, and geography
37	- Lock output format: Markdown, DOCX, slides, or user-provided template
38	- Capture required sections and exact formatting rules
39	- Confirm citation style (footnotes, inline, numbered, APA, etc.)
40	- Confirm length targets per section
41	- Ask for any existing style guide or sample report
42
43	Create a concise report spec file:
44
45	```
46	Report Spec:
47	- Audience:
48	- Purpose:
49	- Scope:
50	- Time Range:
51	- Geography:
52	- Required Sections:
53	- Section Formatting Rules:
54	- Citation Style:
55	- Output Format:
56	- Length Targets:
57	- Tone:
58	- Must-Include Sources:
59	- Must-Exclude Topics:
60	```
61
62	If a user provides a template or an example report, treat it as a hard constraint and mirror the structure.
63
64	### Step 2: Research Plan and Query Set
65
66	Define the research strategy before calling tools:
67
68	- Break the main question into 3-7 subquestions
69	- Define key entities, keywords, and synonyms
70	- Identify primary sources vs secondary sources
71	- Define disqualifiers (outdated, low quality, opinion-only)
72	- Assemble a query set per section
73
74	Use [references/research_plan_checklist.md](references/research_plan_checklist.md) for guidance.
75
76	### Step 3: Evidence Collection (Deepresearch Tool)
77
78	Use the deepresearch tool to collect evidence and citations.
79
80	- Run multiple complete passes if coverage is uncertain
81	- Vary query phrasing to reduce blind spots
82	- Preserve raw tool output in files for traceability
83
84	File structure (recommended):
85	```
86	<output_dir>/research/<topic-name>/
87	deepresearch_pass1.md
88	deepresearch_pass2.md
89	deepresearch_pass3.md
90	```
91
92	If deepresearch is unavailable, rely on user-provided sources only and state limitations explicitly.
93
94	### Step 4: Source Triage and Evidence Table
95
96	Normalize and score sources before drafting:
97
98	- De-duplicate sources across passes
99	- Score sources using [references/source_quality_rubric.md](references/source_quality_rubric.md)
100	- Build an evidence table mapping claims to sources
101
102	Evidence table minimum columns:
103
104	- Source ID
105	- Title
106	- Publisher
107	- Date
108	- URL or reference
109	- Quality tier (A/B/C)
110	- Notes
111
112	### Step 5: Outline and Section Map
113
114	Create an outline that enforces the format contract:
115
116	- Use the template in [references/research_report_template.md](references/research_report_template.md)
117	- Produce a section map with required elements per section
118	- Confirm ordering and headings match the report spec
119
120	### Step 6: Multi-Pass Full Drafting (Parallel Subagents)
121
122	Avoid single-pass drafting; generate multiple complete reports, then merge.
123
124	#### Preferred Strategy: Parallel Subagents (Complete Draft Each)
125
126	Use the Task tool to spawn parallel subagents with isolated context. Each subagent must:
127
128	- Load the report spec, outline, and evidence table
129	- Draft the FULL report (all sections)
130	- Enforce formatting rules and citation style
131
132	Implementation pattern:
133	```
134	Task(subagent_type="general-purpose", prompt="Draft complete report ...", run_in_background=false) -> version1.md
135	Task(subagent_type="general-purpose", prompt="Draft complete report ...", run_in_background=false) -> version2.md
136	Task(subagent_type="general-purpose", prompt="Draft complete report ...", run_in_background=false) -> version3.md
137	```
138
139	Write drafts to files, not conversation context:
140	```
141	<output_dir>/intermediate/<topic-name>/version1.md
142	<output_dir>/intermediate/<topic-name>/version2.md
143	<output_dir>/intermediate/<topic-name>/version3.md
144	```
145
146	### Step

People who liked this also viewed

6 picks

Type