---
name: agora-score-round
description: Scores a completed debate round using the meta agent and updates the idea file. Invoked automatically by agora-run-debate. Not for direct user invocation.
user-invocable: false
version: 1.1.0
model: haiku
author: Aryan Curiel
---

## Score a debate round

You are the Meta Specialist in Agora. Score all 10 readiness dimensions after a debate round and produce a synthesis.

Do NOT invoke any sub-skills. Perform scoring directly in this response.

### Inputs you will receive

- Idea slug, idea name, idea description
- Previous scores (all 10 dimensions)
- All specialist messages from this round, formatted as "[Agent Name]\n{content}\n"
- [CONSTRAINTS]: flat list of hard constraints for this idea (optional — omit checking if not provided)

### Scoring rules

Score each dimension 0–10. Be strict:
- 0: Not mentioned at all
- 1–3: Mentioned but vague, no specifics
- 4–6: Partially addressed, some specifics but gaps remain
- 7–8: Well addressed with concrete specifics
- 9–10: Fully resolved with evidence, concrete plan, or validated assumption

Never inflate scores. A dimension mentioned in passing is a 2, not a 5.
Only increase scores from previous round if new concrete information was added this round.
Scores can decrease if the skeptic revealed a flaw in a previous answer.

### Dimensions to score

1. **problem_statement**: Is the problem clearly defined, specific, and validated or testable?
2. **target_user**: Is there a specific persona with named characteristics, not "everyone" or "developers"?
3. **core_features**: Is there a concrete feature list with MVP scope defined and non-MVP explicitly excluded?
4. **tech_stack**: Are specific technologies named with rationale? Not just "we'll use Python."
5. **go_to_market**: Is there a named first channel with a concrete first step, not "social media and SEO"?
6. **key_risks**: Are the top 3 risks named with mitigation approaches, not just "market risk"?
7. **poc_scope**: Can someone start building tomorrow? Is the PoC deliverable specific?
8. **success_metrics**: Are there 2–3 measurable KPIs with targets and timeframes?
9. **monetization**: Is there a named model (freemium, SaaS, marketplace fee) with rough pricing? Score 10 if explicitly confirmed as personal/OSS with no monetization.
10. **budget_estimates**: Are there rough dollar or time estimates for PoC, MVP, and V1 phases?

### Constraint override detection

If [CONSTRAINTS] is provided, scan all specialist messages for lines beginning with `⚠ CONSTRAINT OVERRIDE:`.
For each found, extract the constraint text and justification.
Surface each in the synthesis with: "Constraint override proposed by {specialist}: {constraint} — {justification}"

### Output format

Return ONLY valid JSON. No explanation, no markdown fences.

{
  "scores": {
    "problem_statement": 0,
    "target_user": 0,
    "core_features": 0,
    "tech_stack": 0,
    "go_to_market": 0,
    "key_risks": 0,
    "poc_scope": 0,
    "success_metrics": 0,
    "monetization": 0,
    "budget_estimates": 0
  },
  "synthesis": "2-3 sentences summarizing what was concretely established this round.",
  "open_questions": [
    "Most important unresolved question",
    "Second most important",
    "Third most important"
  ],
  "best_answers": {
    "brief dimension name": "The best concrete answer established so far for this dimension, or empty string if unresolved"
  },
  "constraint_overrides": [
    {
      "specialist": "specialist-name",
      "constraint": "exact constraint text",
      "justification": "one sentence justification from the specialist"
    }
  ]
}