research

Run structured general-purpose research and produce human-readable reports. Use when users ask for fact-finding, comparisons, market/trend analysis, or evidence-based summaries with quick/standard/deep depth.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "research" with this command: npx skills add alpoxdev/hypercore/alpoxdev-hypercore-research

Research Skill

General research -> human-readable structured report.

<purpose>

Input: topic (natural language) + optional depth (--quick / default / --deep) Output: .hypercore/research/[NN].topic_summary.md

</purpose>

<trigger_conditions>

TriggerAction
/research AI agent framework comparisonTechnical comparison research
/research --deep Korea SaaS market trendsDeep market research
/research --quick WebSocket vs SSEFast technical comparison
"Research X for me"Clarify topic, then execute

If ARGUMENT is missing, ask immediately: "What topic should I research?"

</trigger_conditions>

<depth_levels>

Modequickstandard (default)deep
Query count3-55-1010-15
Agentsresearcher 2 + explore 0-1researcher 3-4 + explore 0-1researcher 4-5 + explore 1 + MCP
Iterative passNoNoYes
Minimum sources51020+
Report size500-1000 chars1500-3000 chars3000-6000 chars

</depth_levels>

<topic_classification>

TypeKeywordsChannels
Technical comparisonvs, compareWebSearch + explore(gh)
Market/trendmarket, trendWebSearch + Firecrawl
Competitor analysiscompetitor, alternativesWebSearch + GitHub MCP
Academic/conceptprinciple, paperWebSearch(arXiv) + WebFetch
Internal projectour codeexplore + Grep
Library/packagepackage@versionDelegate to docs-fetch

</topic_classification>

<mandatory_reasoning>

Mandatory Sequential Thinking

  • Always use sequential-thinking for Phase 1 (query strategy design).
  • For deep mode, also use sequential-thinking in Phase 3 (gap analysis and second-pass query planning).
  • Include current year (2026) in recency-sensitive query sets.
  • Do not produce conclusions without explicit structured reasoning.

</mandatory_reasoning>

<parallel_agent_execution>

  • Use Agent Teams first when 3+ workers are needed.
  • Fallback to parallel Task calls when Agent Teams is unavailable.
  • quick mode (<=2 workers) may run direct parallel tasks.

</parallel_agent_execution>

<workflow>
PhaseTaskTool
0Parse input + detect MCP + classify topicToolSearch
1Build search strategySequential Thinking (2 steps)
2Parallel collectionresearcher + explore + MCP
3Gap analysis + second-pass collection (deep only)analyst -> researcher
4Build reportgeneral-purpose
5Save + return concise summaryWrite

Phase 1 requirements

  • Define 3-5 core research questions
  • Define scope (time/region/language)
  • Generate bilingual query set when useful
  • Assign channels/agents intentionally

Phase 4 writing principles

  • Conclusion first (pyramid principle)
  • Every key claim must include source URL
  • Progressive disclosure (summary -> detail)
  • Use comparison tables where relevant
</workflow>

<report_template>

# [Topic] Research Report

> Date: YYYY-MM-DD | Depth: quick/standard/deep | Sources: N reviewed, M cited

## Executive Summary
[250-400 chars, conclusion first]

## 1. Research Scope
### 1.1 Background
### 1.2 Scope
### 1.3 Method

## 2. Key Findings
### 2.1 [Finding 1]
Core: [one-line summary]
Details: ...
Source: [Title](URL)

## 3. Comparative Analysis (if needed)
| Criteria | A | B | C |
|------|---|---|---|

## 4. Trends and Implications

## 5. Conclusion and Recommendations

## References
- [Title](URL)

</report_template>

<validation>
ItemRequired
ARGUMENTAsk immediately if missing
StrategySequential-thinking trace for query strategy
Sourcesquick 5+, standard 10+, deep 20+
RecencyInclude year/date awareness in source checks
OutputExecutive summary + sources + recommendations
Save.hypercore/research/[NN].*.md

Forbidden:

  • Claims without sources
  • Comparison conclusions without comparison evidence
  • Exiting without saving output
</validation>

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

Coding

crawler

No summary provided by upstream source.

Repository SourceNeeds Review
Coding

elon-musk

No summary provided by upstream source.

Repository SourceNeeds Review
Coding

version-update

No summary provided by upstream source.

Repository SourceNeeds Review