SKILL.md · jhustle44/jcr-guide-pro

Code Commits Issues Pulls Cloud Agent CI/CD

copilot/change-app-name-to-jcrguru

copilot/add-firebase copilot/change-app-name-to-jcrguru copilot/create-zip-file copilot/create-zip-file-again copilot/merge-any-updates main master

md 517 lines 17.6 KB

Rendered Raw

1	---
2	name: deep-research
3	description: Conduct thorough, multi-source research with structured reports and source scoring.
4	---
5
6	# Deep Research
7
8	Conduct comprehensive, multi-source research on complex topics. Systematically gather, evaluate, triangulate, and synthesize information into structured reports with proper citations and source credibility scoring.
9
10	Autonomy Principle: Operate independently. Infer assumptions from context (technical query = technical audience, comparison = balanced perspective, trend = recent 1-2 years). Only stop for critical errors or incomprehensible queries.
11
12	## When to Use
13
14	Activate this skill when the user's request matches any of these patterns:
15
16	### Explicit research requests
17
18	- "Research this," "find out about," "do a deep dive on"
19	- "Write a research report on..."
20
21	- "White paper on..." / "Briefing on..."
22	- "Investigate [topic]"
23
24	#### Industry & market analysis
25
26	- "What's the state of [industry/market]?"
27	- "Survey the landscape of..."
28
29	- "Competitive landscape of [industry]" (no stock tickers involved)
30	- "Trend analysis on [topic]"
31
32	- "How does [country/region] compare for [business activity]?"
33
34	##### Decision-support research
35
36	- "Due diligence on [private company/market]" (non-public companies)
37	- "What should I know before [entering a market / making a decision]?"
38
39	- "Pros and cons of [strategy/approach/technology]"
40	- "What are the risks of [strategy/market/decision]?"
41
42	- "Benchmark [X] against [industry/peers]" (non-financial benchmarking)
43
44	###### Verification & evidence-based analysis
45
46	- "Fact-check this" / "Is [claim] true?"
47	- "What does the research say about..."
48
49	- Needs to verify claims or compare conflicting information
50	- Wants to understand a topic from multiple angles with cited sources
51
52	###### Academic & technical evaluation
53
54	- Literature review, technology evaluation, state-of-the-art survey
55
56	## When NOT to Use
57
58	- Simple factual lookups (1-2 searches) --> use web-search directly
59	- Searching within the user's codebase --> use grep/glob
60
61	- Replit-specific features --> use replit-docs skill
62	- Specific stock tickers, public company financials, DCF models, portfolio analysis --> use stock-analyzer (it calls deep-research internally for web research)
63
64	- "Analyze this dataset" with user-provided data --> use data-visualization
65	- "Build a presentation on..." --> use slides skill
66
67	## Decision Tree
68
69	```text
70
71	Request Analysis
72
73	+-- Simple factual lookup (1-2 searches)? --> STOP: Use web-search directly
74
75	+-- Searching user's codebase? --> STOP: Use grep/glob
76
77	+-- Replit-specific feature question? --> STOP: Use replit-docs skill
78
79	+-- Mentions stock tickers or public company --> STOP: Use stock-analyzer
80
81	financials/valuations/DCF? (it calls deep-research internally)
82
83	+-- Wants to analyze a user-provided dataset? --> STOP: Use data-visualization
84
85	+-- Complex multi-angle analysis needed? --> CONTINUE to Depth Selection
86
87	```
88
89	## Depth Selection
90
91	Select the research depth based on the request complexity. Default to standard when unclear.
92
93	\| Tier \| Subagents \| Min Sources \| Phases \| Estimated Time \| When to Use \|
94
95	\|---\|---\|---\|---\|---\|---\|
96
97	\| Quick \| 3 \| 8 \| 1, 2, 3, 7 \| 2-5 min \| Focused question, single domain, time-sensitive \|
98
99	\| Standard \| 5 \| 15 \| 1-7 \| 5-15 min \| Most research requests, market analysis, technology evaluation \|
100
101	\| Deep \| 5 + 2 gap-fill \| 25 \| 1-8 (all) \| 15-30 min \| Critical decisions, comprehensive reviews, multi-stakeholder analysis \|
102
103	### Selection heuristics
104
105	- User says "quick overview" or "brief research" --> Quick
106	- User says "research this" or "analyze" --> Standard
107
108	- User says "deep dive," "comprehensive," or "exhaustive" --> Deep
109	- Country/industry analysis, investment research --> Standard or Deep
110
111	- Technology comparison with 3+ options --> Standard
112	- Literature review, state-of-the-art survey --> Deep
113
114	## Methodology
115
116	### Phase 1: Scope Definition
117
118	Before starting research, clearly define:
119
120	- Research question: What specific question(s) are you answering?
121	- Scope boundaries: What is in/out of scope?
122
123	- Depth tier: Quick, Standard, or Deep (see table above)
124	- Audience: Technical, executive, general? (Infer from context if not stated)
125
126	- Output expectations: Report format, approximate length
127
128	Then run 1-2 broad landscape searches to orient yourself and identify the focus areas for decomposition.
129
130	### Phase 2: Planning & Decomposition
131
132	Decompose the topic into distinct, non-overlapping focus areas based on the depth tier:
133
134	- Quick: 3 focus areas
135	- Standard: 5 focus areas
136
137	- Deep: 5 focus areas (gap-fill subagents added later in Phase 5)
138
139	How to decompose: After the broad landscape search, identify angles that together cover the full topic without significant overlap. For example, researching "state of electric vehicles 2026" might decompose into:
140
141	1. Market & Competition -- market share, sales figures, manufacturer rankings
142	2. Technology -- battery chemistry, charging standards, range improvements
143
144	3. Policy & Regulation -- government incentives, emissions mandates, trade tariffs
145	4. Infrastructure -- charging network growth, grid capacity, urban vs rural
146
147	5. Consumer & Economics -- total cost of ownership, resale value, adoption demographics
148
149	### Phase 3: Parallel Source Discovery via Subagents
150
151	Launch all research subagents simultaneously using `startAsyncSubagent`. Each subagent gets a specific focus area, tailored search terms, and a structured output template.
152
153	#### Subagent task template
154
155	```javascript
156
157	await startAsyncSubagent({
158
159	task: `Research FOCUS AREA: [Area Name]
160
161	Topic context: [1-2 sentence description of the overall research question]
162
163	Your job: Search for information specifically about [focus area]. Run at least 4 webSearch queries with different angles:
164
165	- [specific search term 1]
166	- [specific search term 2]
167
168	- [specific search term 3]
169	- [specific search term 4]
170
171	IMPORTANT - Local language searches: If this topic is specific to a non-English-speaking country or region, run at least 1-2 searches in the local language (e.g., Spanish for Latin America, Portuguese for Brazil, French for Francophone Africa). Primary government data, local media, and industry reports are often only available in the local language.
172
173	IMPORTANT - Source freshness: Note the publication date of every source. Flag any source older than 18 months on a fast-moving topic.
174
175	For the most promising results, use webFetch to read the full article. If webFetch returns empty content, try an alternative URL from the search results or run a more specific search query.
176
177	Return your findings using this EXACT structure:
178
179	## Key Facts
180
181	[Bullet list of key data points, each with source URL in parentheses]
182
183	## Notable Claims Requiring Cross-Reference
184
185	[Claims that seem important but only appear in one source, or that conflict with other findings]
186
187	## Source Quality Assessment
188
189	[For each source, rate: Tier 1 (government/multilateral/academic), Tier 2 (major publication/industry report), Tier 3 (blog/opinion/promotional). Note publication date.]
190
191	## Gaps & Unanswered Questions
192
193	[What you could not find or what needs deeper investigation]
194
195	## Sources
196
197	[Numbered list with title, URL, publication date (if available), and tier rating]
198
199	Minimum: 5 distinct sources with URLs`
200
201	});
202
203	```
204
205	##### Launch pattern
206
207	```javascript
208
209	// Launch all subagents simultaneously
210
211	await startAsyncSubagent({ task: `Research FOCUS AREA 1: [Area] ...` });
212
213	await startAsyncSubagent({ task: `Research FOCUS AREA 2: [Area] ...` });
214
215	await startAsyncSubagent({ task: `Research FOCUS AREA 3: [Area] ...` });
216
217	// ... (3 for Quick, 5 for Standard/Deep)
218
219	```
220
221	Then call the `wait_for_background_tasks` tool to collect their results.
222
223	Each subagent should:
224
225	- Run 4+ `webSearch` queries with different phrasings and angles
226	- Include at least 1-2 local-language searches for country/region-specific topics
227
228	- Use `webFetch` on the 2-3 most relevant results to extract detailed data
229	- Retry with alternative URLs if webFetch returns empty content
230
231	- Return findings using the structured template above
232	- Flag any claims that conflict with other results
233
234	- Note publication dates and assess source freshness
235
236	This approach gathers 25+ distinct sources across focus areas simultaneously, producing far more comprehensive coverage than sequential searching.
237
238	### Phase 4: Triangulation & Source Evaluation
239
240	After collecting all subagent results, systematically evaluate and cross-reference:
241
242	#### Source credibility tiers
243
244	\| Tier \| Description \| Examples \| Weight \|
245
246	\|---\|---\|---\|---\|
247
248	\| Tier 1 \| Government, multilateral, academic, official statistics \| IMF, World Bank, central banks, peer-reviewed journals, SEC filings \| Highest -- treat as ground truth unless contradicted by multiple Tier 1 sources \|
249
250	\| Tier 2 \| Major publications, established industry reports, reputable news \| Reuters, Bloomberg, Chambers & Partners, LAVCA, Big 4 reports \| High -- reliable but verify key claims \|
251
252	\| Tier 3 \| Industry blogs, company websites, opinion pieces, promotional content \| Company press releases, consultant blogs, sponsored content \| Supporting only -- never use as sole source for a claim \|
253
254	##### Triangulation rules
255
256	- Every major claim must be supported by 3+ sources (at least 2 Tier 1 or Tier 2)
257	- When sources conflict on a data point, prefer: Tier 1 > Tier 2 > Tier 3, and more recent > older
258
259	- When Tier 1 sources conflict with each other, present both figures and note the discrepancy
260	- Flag any finding that rests on a single source, regardless of tier
261
262	- Note the publication date of key data points; flag anything older than 18 months on fast-moving topics
263
264	###### Conflict resolution for quantitative data
265
266	When multiple sources report different numbers (e.g., GDP figures, market sizes):
267
268	1. Prefer the primary/official source (e.g., central bank over news article)
269	2. If both are primary sources, present the range and note the methodology difference
270
271	3. Never silently pick one number -- acknowledge the variance
272
273	### Phase 5: Gap Analysis & Follow-Up (Standard and Deep tiers only)
274
275	Review the collected findings for completeness:
276
277	- Identify claims with fewer than 3 supporting sources
278	- Identify sections with thin coverage or missing data
279
280	- Note unanswered questions flagged by subagents
281	- Check whether any focus area returned significantly fewer sources than others
282
283	For Deep tier only: Launch 1-2 targeted follow-up subagents to fill the most critical gaps:
284
285	```javascript
286
287	await startAsyncSubagent({
288
289	task: `GAP-FILL RESEARCH: [Specific gap identified]
290
291	Context: During initial research on [topic], we found insufficient data on [gap].
292
293	Your job: Run 3-4 targeted searches to fill this specific gap:
294
295	- [targeted search term 1]
296	- [targeted search term 2]
297
298	- [targeted search term 3]
299
300	Return findings using the same structured template as the initial research.`
301
302	});
303
304	```
305
306	### Phase 6: Critique & Self-Review (Deep tier only)
307
308	Before writing the final report, conduct a critical self-review:
309
310	- Weak claims: Are any findings supported by only Tier 3 sources? Downgrade or remove them.
311	- Balance: Does the report present multiple perspectives, or does it lean toward one viewpoint?
312
313	- Logical coherence: Do the findings tell a consistent story? Are there contradictions that need to be addressed?
314	- Completeness: Does the report answer the original research question fully?
315
316	- Freshness: Are key data points current, or are they based on outdated sources?
317	- Speculation vs. fact: Is every claim clearly labeled as established fact, expert opinion, or speculation?
318
319	### Phase 7: Report Writing & Synthesis
320
321	Write the report following these principles:
322
323	#### Writing style
324
325	- Prose-first (80%+ prose). Write in flowing paragraphs, not bullet-point dumps. Bullets are acceptable for data tables, source lists, and comparison matrices -- but the analysis itself should be narrative.
326	- Lead with the most important findings
327
328	- Support every factual claim with an inline citation [N] referencing the numbered source list
329	- Acknowledge limitations and uncertainties explicitly
330
331	- Distinguish clearly between established facts, expert opinions, and speculation
332	- Provide actionable recommendations where appropriate
333
334	##### Section word count guidance
335
336	\| Section \| Target Length \|
337
338	\|---\|---\|
339
340	\| Executive Summary \| 200-400 words \|
341
342	\| Background/Context \| 200-500 words \|
343
344	\| Each Major Finding \| 400-1,500 words \|
345
346	\| Analysis/Synthesis \| 500-1,000 words \|
347
348	\| Limitations \| 100-300 words \|
349
350	\| Recommendations \| 200-500 words \|
351
352	###### Quality gates (must pass before delivering)
353
354	\| Gate \| Requirement \|
355
356	\|---\|---\|
357
358	\| Source count \| Quick: 8+, Standard: 15+, Deep: 25+ \|
359
360	\| Claims per finding \| 3+ sources supporting each major claim \|
361
362	\| Citation coverage \| Every factual claim has an inline [N] citation \|
363
364	\| No placeholders \| Zero "TBD," "to be determined," or "[need source]" entries \|
365
366	\| Source list complete \| Every [N] citation maps to a numbered source with URL \|
367
368	\| Freshness \| Key data points are from the last 18 months (flagged if older) \|
369
370	\| Prose ratio \| 80%+ of analysis sections are prose, not bullets \|
371
372	## Output Format
373
374	### Research Report Structure
375
376	```text
377
378	# [Research Topic]
379
380	Research Date: [Date]
381
382	Depth: [Quick / Standard / Deep]
383
384	Sources Consulted: [Number]
385
386	## Executive Summary
387
388	[2-3 paragraph overview of key findings and conclusions. 200-400 words.]
389
390	## Background
391
392	[Context needed to understand the topic. 200-500 words.]
393
394	## Key Findings
395
396	### Finding 1: [Theme]
397
398	[Detailed prose analysis with inline source citations [N]. 400-1,500 words.]
399
400	### Finding 2: [Theme]
401
402	[Detailed prose analysis with inline source citations [N]. 400-1,500 words.]
403
404	### Finding 3: [Theme]
405
406	[Detailed prose analysis with inline source citations [N]. 400-1,500 words.]
407
408	[Additional findings as needed -- typically 4-8 for Standard/Deep]
409
410	## Analysis
411
412	[Cross-cutting analysis, patterns, implications. Draw connections between
413
414	findings that reveal insights not visible in any single section. 500-1,000 words.]
415
416	## Limitations
417
418	[What couldn't be determined, data gaps, source constraints, caveats. 100-300 words.]
419
420	## Recommendations
421
422	[Actionable next steps based on findings. 200-500 words.]
423
424	## Sources
425
426	[Numbered list of all sources with: title, URL, publication date, tier rating]
427
428	1. [Title] -- [URL] (Published: [date], Tier [1/2/3])
429	2. ...
430
431	```
432
433	## Example Workflow
434
435	```javascript
436
437	// Phase 1: Scope -- broad landscape search
438
439	const overview = await webSearch({ query: "state of electric vehicle market 2026" });
440
441	// Determine depth tier (Standard for this example)
442
443	// Identify 5 focus areas from overview results
444
445	// Phase 3: Launch 5 parallel research subagents
446
447	await startAsyncSubagent({
448
449	task: `Research FOCUS AREA 1: EV Market & Competition
450
451	Topic context: Comprehensive analysis of the global electric vehicle industry in 2026.
452
453	Your job: Search for information specifically about EV market share, sales figures, and manufacturer rankings. Run at least 4 webSearch queries:
454
455	- "EV market share by manufacturer 2025 2026"
456	- "electric vehicle sales global rankings"
457
458	- "Tesla BYD market share comparison 2026"
459	- "electric vehicle market size revenue 2026"
460
461	For country-specific angles, also search in relevant local languages.
462
463	If webFetch returns empty content, try alternative URLs.
464
465	Return findings using the structured template with Key Facts, Notable Claims, Source Quality Assessment, Gaps, and Sources.`
466
467	});
468
469	await startAsyncSubagent({ task: `Research FOCUS AREA 2: EV Battery Technology ...` });
470
471	await startAsyncSubagent({ task: `Research FOCUS AREA 3: EV Policy & Regulation ...` });
472
473	await startAsyncSubagent({ task: `Research FOCUS AREA 4: EV Charging Infrastructure ...` });
474
475	await startAsyncSubagent({ task: `Research FOCUS AREA 5: EV Consumer Economics ...` });
476
477	// Phase 4: Triangulate -- evaluate sources, resolve conflicts, score credibility
478
479	// Phase 5: Gap analysis -- identify weak spots, launch follow-up if Deep tier
480
481	// Phase 6: Critique -- self-review for Deep tier
482
483	// Phase 7: Write the report following quality gates and prose-first style
484
485	// Save to research/[topic].md and present to user
486
487	```
488
489	After the launch block above, call the `wait_for_background_tasks` tool to wait for subagents, then proceed to Phase 4.
490
491	## Best Practices
492
493	1. Cast a wide net first, then narrow -- start with broad searches before diving into specifics
494	2. Cross-reference critical claims -- never rely on a single source for important facts; require 3+ sources for major claims
495
496	3. Cite everything inline -- every factual claim should have a [N] citation immediately following it
497	4. Note disagreements -- when sources conflict, present both sides and analyze why
498
499	5. Timestamp your research -- note when the research was conducted, as information changes
500	6. Separate facts from analysis -- clearly distinguish between what sources say and your interpretation
501
502	7. Search in local languages -- for country/region-specific research, primary sources are often only in the local language
503	8. Handle fetch failures gracefully -- if webFetch returns empty content, try alternative URLs or run more targeted search queries
504
505	9. Prefer authoritative sources -- when numbers conflict, prefer Tier 1 (government/multilateral) over Tier 2 (publications) over Tier 3 (blogs/promotional)
506	10. Flag stale data -- note when key data points are older than 18 months on fast-moving topics
507
508	## Limitations
509
510	- Cannot access paywalled academic journals or subscription databases
511	- Cannot access social media content (LinkedIn, Twitter, Reddit)
512
513	- Web sources may have varying levels of reliability
514	- Research is a snapshot in time -- findings may change
515
516	- Cannot conduct primary research (surveys, interviews, experiments)
517	- webFetch may fail on some pages (JavaScript-heavy sites, paywalled content); mitigate with retry and alternative URLs