Skip to content
Moira Documentation

Iterative Research

Research workflow with iterative quality improvement through critique/improve cycles. A critic subagent evaluates results, suggests improvements, and the process repeats until quality gates are met.

Start

Section titled “Start”
Terminal window
mcp__moira__start({ workflowId: "moira/iterative-research", parentExecutionId: "none" })

Process

Section titled “Process”
flowchart LR
    A[Setup] --> B[Generate]
    B --> C[Critique]
    C --> D{Quality OK?}
    D -->|No| E[Improve]
    E --> C
    D -->|Yes| F[Publish]

Steps

Section titled “Steps”
StepActionOutput
1. SetupConfigure research parameters and workspaceResearch config
2. GenerateGenerate initial researchDraft research
3. CritiqueQuality assessment with metricsCritique report
4. Quality CheckEvaluate against quality gatesPass/Fail decision
5. ImproveAddress critique feedbackImproved research
6. PublishPrepare and publish final versionPublished research

Features

Section titled “Features”

Depth Levels

Section titled “Depth Levels”
LevelSourcesWordsFocus
quick5-81500-2500Rapid overview
normal12-203000-5000Practical guidance
deep25-356000-10000Strategic analysis
scientific40+10000+Academic rigor

Quality Gates

Section titled “Quality Gates”
  • Critical issues: 0
  • Major issues: 0
  • Formatting score: ≥ 8/10

Iteration Limits

Section titled “Iteration Limits”
  • Maximum 5 iterations
  • Force completion option if limit reached
  • User decision on forced completion

Issue Classification

Section titled “Issue Classification”
SeverityDescription
CriticalFundamental flaws, factual inaccuracies
MajorClarity problems, source attribution gaps
ModerateDepth enhancement, alternative perspectives
MinorPolish, formatting consistency

Example Node Configuration

Section titled “Example Node Configuration”
{
  "id": "critique-research",
  "type": "agent-directive",
  "directive": "Assess research quality using {{research_critique_prompt}}",
  "completionCondition": "Critique completed with numerical metrics",
  "inputSchema": {
    "type": "object",
    "properties": {
      "critical_issues_count": { "type": "number" },
      "major_issues_count": { "type": "number" },
      "formatting_score": { "type": "number" }
    }
  }
}

Difference from Verified Research

Section titled “Difference from Verified Research”
WorkflowFocus
moira/verified-researchLinear 8-step, source verification
moira/iterative-researchIterative with quality improvement cycles

Use iterative-research when you need high-quality output with multiple revision cycles.

Section titled “Related”