Skip to main content Home Creators javimosch open-claw-skills research-reprompter
research-reprompter Transform rough research questions into executable USACF research prompts.
Use when user says "research", "research this", "investigate", "deep dive",
"researcher", or pastes a research topic. Generates complete multi-agent swarm
configuration with algorithm selection, claude-flow commands, and adversarial review.
Jump to install Skills Marketplace Discover and explore AI skills built by the community.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Show prompt details A direct command skips the review prompt. Inspect the source before running it.
npx skills add https://github.com/javimosch/open-claw-skills --skill research-reprompterThe command stays on one line. Scroll horizontally to inspect it before copying.
Prefer a local copy? Download the files currently available to SkillsMP.
Download Zip Downloading... More from this repository
Related occupations SOC
Based on SOC occupation classification
name research-reprompter description Transform rough research questions into executable USACF research prompts.
Use when user says "research", "research this", "investigate", "deep dive",
"researcher", or pastes a research topic. Generates complete multi-agent swarm
configuration with algorithm selection, claude-flow commands, and adversarial review.
compatibility Full features require Claude Code with claude-flow installed (npx claude-flow@alpha).
Core prompt generation works on all Claude surfaces.
metadata {"version":"2.0.0"}
Researcher v2.0 (USACF Research Generator)
Voice-to-research engineering for Claude Code. Transform rough questions into executable USACF swarm configurations.
Changelog
Version Changes v2.0 Full USACF integration: algorithm selection, claude-flow commands, adversarial review, fact-checking, memory namespaces v1.0 Initial version based on Reprompter v4.1
Purpose
Turn your rough research questions into complete, executable multi-agent research prompts using the USACF framework.
The Problem:
Research questions are often vague and unstructured
Manual setup of research swarms is tedious
Missing adversarial review leads to blind spots
No systematic algorithm selection The Solution:
Smart interview → USACF super-prompt with all phases, agents, and commands.
Process (4 steps)
Step 1: Receive raw input Accept the user's rough research question (dictated, typed messily, or incomplete).
Trigger words: research, investigate, deep dive, analyze, research this
Step 2: Complexity detection Auto-detect complexity to select algorithm:
Simple (< 20 words, single topic) → CoT, 1-3 agents
Medium (branching, comparison) → ToT, 4-8 agents
Complex (comprehensive, multi-domain) → GoT, 9-15 agents
Step 3: Smart interview (gather user input)
Research Title - Name for this researchSubject - What we're researchingSubject Type - Product / Software / Business / Process / OrganizationResearch Type - Competitive / Gap / Technical / Due Diligence / MarketObjectives - What to find out (1-20, one per line)Constraints - Focus areas, limitations (optional)Depth - CoT / ToT / GoTOutput - Brief / Full Report / Action Plan / Raw
Step 4: Generate USACF super-prompt + score Generate complete executable configuration with:
Initialization commands
All phase agents (Discovery, Analysis, Adversarial, Synthesis)
Memory operations
Final report generator
Quality score comparison
CRITICAL: MUST GENERATE COMPLETE SUPER-PROMPT After interview completes, you MUST immediately:
Select algorithm (CoT/ToT/GoT) based on complexityGenerate full USACF super-prompt with ALL phasesInclude claude-flow commands for every operationAdd adversarial review agents (red-team, fact-checker)Show quality score (before/after comparison)Offer to execute or copy WRONG: Generate simple prompt without agents
RIGHT: Generate full USACF config with all phases, agents, memory ops
Algorithm selection matrix Complexity Algorithm Topology Agents When to Use Simple Chain-of-Thought (CoT) Star 1-3 "What is X?" Single topic Medium Tree-of-Thought (ToT) Hierarchical 4-8 "Compare X vs Y" Branching Complex Graph-of-Thought (GoT) Mesh/Hive 9-15+ "Comprehensive analysis"
Simple: Single topic, factual question, < 20 words
Medium: "compare", "vs", "evaluate", "options"
Complex: "comprehensive", "gaps and opportunities", multiple domains
USACF phases (all required for complex research)
Phase 0: Initialization npx claude-flow@alpha init --force
npx claude-flow@alpha swarm init --topology {topology} --max-agents {N}
npx claude-flow@alpha memory store "session/config" '{...}' --namespace search
Phase 0.5: Meta-analysis
Step-back prompting (principles, criteria)
Self-ask decomposition (15-20 questions)
Research planning (ReWOO)
Phase 1: Discovery (Parallel)
component-identifier
hierarchy-analyzer
interface-mapper
flow-tracer
Phase 2: Analysis (Parallel)
6 gap hunters (quality, performance, security, structural, capability, UX)
4 risk analysts (FMEA, edge cases, vulnerabilities, reliability)
Phase 2.5: Adversarial review (critical)
red-team-reviewer : Challenge ALL findingsfact-checker : RAG verification with web_searchcoordinator : Integrate feedback, update confidence
Phase 3: Synthesis (Parallel)
quick-win-generator (0-3 months)
strategic-generator (3-12 months)
transformational-generator (12-36 months)
pareto-optimizer (multi-objective portfolios)
Phase 4: Final report
Ultra-brief (3 sentences)
Executive summary
Top 10 findings with confidence
Recommended actions by horizon
Limitations & uncertainties
Memory namespace convention All agents store to namespaced memory:
session/config
meta/principles
meta/questions
meta/research-plan
discovery/components
discovery/hierarchy
discovery/interfaces
discovery/flows
gaps/quality
gaps/performance
gaps/security
risks/fmea
risks/edge-cases
risks/vulnerabilities
adversarial/critiques
adversarial/fact-check
opportunities/quick-wins
opportunities/strategic
opportunities/transformational
opportunities/pareto-recommendation
output/final-report
Quality scoring Always show before/after metrics:
Dimension Before After Change Clarity X/10 X/10 +X% Algorithm Selection 0/10 10/10 +∞ Agent Design 0/10 9/10 +∞ Memory Ops 0/10 10/10 +∞ Adversarial 0/10 9/10 +∞ Fact Checking 0/10 8/10 +∞ Overall X/10 9+/10 +2000%+
Example
"look into what solana is doing with AI and how we compare"
After (USACF super-prompt):
# USACF Research: Solana AI Competitive Analysis
## Configuration
- Algorithm: ToT (medium complexity - comparison)
- Topology: Hierarchical
- Agents: 8
- Output: Executive Brief
## Phase 0: Initialization
[claude-flow init commands]
## Phase 1: Discovery
[4 parallel agents with memory stores]
## Phase 2: Analysis
[Gap hunters + risk analysts]
## Phase 2.5: Adversarial
[Red team + fact checker]
## Phase 3: Synthesis
[Opportunity generators + pareto optimizer]
## Phase 4: Report
[Final report generator]
Quality: 1.2/10 → 9.3/10 (+675%)
Tips for best results
Be specific about competitors - Name them in the inputMention constraints early - "focus on Q1", "executive-level"State objectives - Even rough ones helpSay "expand" - For full interview on simple queriesSay "quick" - To skip interview for simple research
Comparison: Reprompter vs Researcher Aspect Reprompter Researcher Trigger "reprompt" "research" / "researcher" Purpose General prompts Research prompts Output Structured prompt USACF swarm config Agents None 8-15 parallel agents Memory No Full namespace system Adversarial No Red team + fact checker Algorithm No CoT/ToT/GoT selection
