skills.rest
PsChina
PsChinaCOMMUNITY
Explore all skills by this author →
GitHub

deep-research

Conducts multi-source deep research with parallel sub-agents and delivers consultant-grade reports.

Updated Jun 1, 2026
Search Tags:#claude code#report generation#multi-agent#competitive analysis#due diligence#deep research#fact verification
One-click install
npx skills add https://github.com/PsChina/deep-research --skill deep-research-pschina
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: deep-research
Source: https://github.com/PsChina/deep-research
Command: npx skills add https://github.com/PsChina/deep-research --skill deep-research-pschina

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Plain Q&A often produces answers that sound right but lack verification. This Skill performs structured deep research: it forms hypotheses, dispatches independent researcher sub-agents to verify in parallel, actively hunts for counter-evidence, checks source recency, and delivers a report with a judgment, confidence level, and traceable sources. ## Core Features & Use Cases - Four-phase recursive research: curiosity loop for broad exploration, deep reasoning with blind-spot detection, adversarial peer review by independent sub-agents, and iterative revision before report delivery. - Quality tiers: fast (5-10 min), standard (15-30 min, default), and deep (40-60 min) modes with explicit search budgets, parallelism limits, and honest degraded flagging when constraints are unmet. - Freshness and bias gates: mandatory freshness sweeps for newly released entities, per-entity successor probing, and systematic bias checks across geography, language, open-source vs commercial, and benchmark sources. - Use Case: Ask it to compare AI coding assistants for a purchase decision; it will map the landscape, cross-verify claims across independent sources, surface dissenting views, and deliver a decision memo with a central thesis and falsifiable conditions. ## Quick Start Say "deep research <topic>" or run /deep-research <topic> in Claude Code, optionally adding --quality=fast, standard, or deep.

Frequently Asked Questions about deep-research

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a deep research task in Claude Code?▼

Say "deep research <topic>" or use the /deep-research <topic> command after copying the skill into ~/.claude/skills/. You can append --quality=fast, standard, or deep to control depth, and add "start now" to skip the 60-second plan confirmation window.

What is the difference between fast, standard, and deep research tiers?▼

Fast takes 5-10 minutes with the main agent only and produces 1000-2500 words. Standard (default) takes 15-30 minutes with at least 2 researchers and 2500-5000 words. Deep takes 40-60 minutes with at least 4 researchers and 5000-12000 words.

When should I not use deep research?▼

Skip it for single-point factual questions, code writing, pure computation, and simple how-to tasks that need no external sources. The skill is designed for open questions requiring cross-verification across multiple authoritative sources, such as selection, comparison, and investigation.

How does deep research avoid fabricated or outdated information?▼

Every claim requires a traceable source URL and date, and mandatory freshness gates verify data recency and probe for successor versions of known entities. Independent peer-review sub-agents attack weak claims, and unverifiable items are explicitly flagged with lowered confidence.

Can deep research handle Chinese-language or regional topics?▼

Yes. Topics involving China, A-shares, or domestic policy trigger bilingual search rounds, and non-English-native entities are searched in their native language. The bias-detection rules require independent investigation of any ecosystem with two or more significant entities.

Related Skills

Explore More Education & Research Skills→
silvio-lsilvio-l

deep-research

Produces multi-phase research reports with sourced, scored evidence and recommendations.

Community
Advanced
ekcheungAIekcheungAI

stanley-druckenmiller-perspective

Analyze global macro opportunities with thesis, evidence, loss limits, and position sizing.

Community
Advanced
adriaanmostert1976-labadriaanmostert1976-lab

fact-checker

Verify factual claims against primary and authoritative web sources.

Community
Advanced
moltis-orgmoltis-org
2.8k

gemini

Generate Q&A, summaries, and content from single prompts via Gemini CLI.

Official
Intermediate