/research - Intelligent Research Orchestration

I'll coordinate specialized agents to conduct systematic research on your topic.

Research Topic

Query: $ARGUMENTS

🤖 Orchestration Mode

I'll analyze your query and build an execution plan using specialized agents.

Agent Registry:

• literature-reviewer : Academic source search, citation chains

• critical-analyzer : Fallacy detection, bias identification

• hypothesis-explorer : Hypothesis formulation, variable mapping

• lateral-thinker : Cross-domain analogies, creative thinking

• qual-researcher : Thematic analysis, coding strategies

• quant-analyst : Statistical methods, effect sizes

• peer-reviewer : Manuscript evaluation

• ethics-expert : IRB compliance, privacy risks

• grant-writer : Grant proposal development, funding strategy

Orchestration Process:

• Classify Query: Determine research type

• Select Agents: Choose optimal agent sequence

• Generate Plan: Build execution DAG

• Present Plan: Show workflow to user

• Execute: Run agents sequentially, save outputs

• Synthesize: Integrate outputs, save final report

I'm analyzing your research topic to determine the optimal workflow...

Proposed Plan: [I will generate this based on your query]

Proceed? (yes/no/modify)

File Writing Protocol

CRITICAL: After plan approval, you MUST save all research outputs to files.

Step 1: Create Output Directory

After user approves the plan, immediately:

• Use Bash to create timestamped directory: mkdir -p research-outputs/$(date +%Y-%m-%d_%H-%M-%S)

• Store the full path in a variable for subsequent file writes

• Confirm directory creation to user

Step 2: Write Research Plan

Use Write tool to create 00-research-plan.md in the output directory:

Research Plan: [Query]

Created: [Timestamp] Query: [User's research query]

Selected Agents

1. [agent-name] - [purpose]
2. [agent-name] - [purpose] ...

Execution Plan

[Full plan as presented to user]

Step 3: Write Agent Outputs

After EACH agent completes execution:

• Use Write tool to save output to {NN}-{agent-name}.md

• NN = sequential number (01, 02, 03, etc.)

• Example: 01-literature-reviewer.md , 02-critical-analyzer.md

• Format:

{Agent Name} Output

Agent: {agent-name} Executed: [Timestamp]

[Full agent output - preserve all markdown formatting]

Step 4: Write Final Synthesis

After synthesis completes:

• Use Write tool to save to final-synthesis.md

• Include complete synthesis with all sections

• Add metadata header with timestamp and agents used

Error Handling

• If directory creation fails: warn user and continue with conversation-only output

• If Write fails: log error, notify user, continue execution

• Partial results are acceptable if execution is interrupted

Modes Supported:

• Default: Interactive

• --auto : Automatic execution

• --plan-only : Show plan only

• --manual : Traditional guided mode

Templates: --template=quick|rigorous|comprehensive