Insight

State Management

• revisionsAttempted: counter, starts at 0, increments on each REVISE cycle
• Inputs (gap, evidence, knowledge) may be refined between revision cycles based on tactic feedback
• Does not maintain papersRead or urlsVisited — those are managed by the calling gap-analysis strategy

Completeness Probe — Insight

5 Whys depth:

• Did the root-cause drilling reach a fundamental cause (not just a symptom)?
• Are there at least 3 levels of "why" in the chain?
• Does the root cause suggest a clear direction for solutions?

Stakeholder coverage:

• Were at least 3 different stakeholder perspectives considered? (e.g., researchers, practitioners, reviewers, end-users)
• Did stakeholder mapping reveal any surprising or non-obvious needs?

Tension mining:

• Were genuine opposing forces identified (not just restated problem)?
• Do the tensions explain WHY this gap persists despite being known?

HMW quality:

• Are the How Might We questions actionable (not too broad, not too narrow)?
• Do they suggest multiple possible solution directions?

If ANY checkbox fails → trigger one REVISE cycle with targeted instructions. This probe runs BEFORE the existing verdict check, not after. </HARD-GATE>

Output

{
  "insight": {
    "rootCause": "Deep root cause of the gap...",
    "stakeholderNeeds": ["need1", "need2"],
    "coreContradiction": "The fundamental tension or conflict...",
    "impliedSolution": "What closing this gap would require..."
  },
  "verdict": "PASS | FAIL",
  "revisionsAttempted": 1,
  "quality": "high | medium | low"
}