Evaluate Idea

Evaluate Idea

Score a business idea by dispatching evaluation agents that research and rate the idea across investment and market dimensions. Produces a structured, machine-readable report.

Protocol

Follow these phases in order.

Phase 1: Load

1. Parse <argument> for the idea folder name and an optional agent filter (vc or market)
2. Search for the idea folder:

ideas/*{idea-folder-name}*/

3. Read all documents in the folder (00-overview through 05-experiments). Note which docs exist and which are missing.
4. Read the shared evaluation framework:

references/evaluation-framework.md

5. Read the Exa research guide (if it exists):

references/exa-research.md

6. If 03-assumptions.md exists, flag it for cross-referencing.

Phase 2: Dispatch Agents

Prepare the context block for agents — combine:

• The full evaluation framework
• The Exa research guide (if loaded)
• All idea documents content
• The assumptions document (if it exists)

If an agent filter was specified (vc or market), run only that agent.

Otherwise, dispatch both agents in parallel:

Launch two Agent tasks simultaneously, each receiving the full context block:

1. VC Agent — evaluates investability across 8 weighted dimensions (Team, Timing, TAM, Technology/Product, Competition/Moat, Business Model, GTM, Traction/Validation)
2. Market Analyst Agent — evaluates market opportunity across 5 dimensions (Market Size & Growth, Competitive Landscape, Timing & Tailwinds, Customer Accessibility, Regulatory/Macro Risk)
3. YC Founder-Fit Agent — evaluates founder-idea fit across 10 dimensions from the YC framework (Founder-Market Fit, Market Size, Problem Acuteness, Competition Presence, Personal/Peer Demand, Recent Possibility/Necessity, Successful Proxies, Long-term Commitment, Business Scalability, Idea Space Fertility)

Each agent:

• Applies the reasoning tools from the evaluation framework
• Uses WebSearch to validate empirical claims
• Scores each dimension 1-5 using the framework's scale and output format
• Returns dimension scores, overall score, and summary

Phase 3: Synthesize Report

Collect outputs from both agents (or one if filtered). Compute:

If all agents ran:

combined_score = round((vc_overall + market_overall + yc_overall) / 3)

If two agents ran: average those two scores. If single agent: use that agent's overall score as the combined score.

Identify:

• Deal-breakers from the VC agent (Team=1, TAM=1, or any two dimensions=1)
• Weakest dimension across all agents — the single lowest-scoring dimension
• Assumptions gaps from the Market Analyst's cross-reference against 03-assumptions.md

Present a summary table:

## Evaluation Summary

| Agent | Score | Deal-breakers |
|-------|-------|---------------|
| VC Investability | [score]/100 | [list or None] |
| Market Opportunity | [score]/100 | — |
| YC Founder-Fit | [score]/100 | — |
| **Combined** | **[score]/100** | **[count] deal-breaker(s)** |

**Weakest dimension:** [agent] → [dimension] ([score]/5) — address this first.

Then present the full per-dimension analysis from each agent.

Phase 4: Write Evaluation File

Write the evaluation report to the idea folder:

Filename: evaluation-YYYYMMDD-HHmmss.md Location: Inside the idea folder (ideas/{idea-name}/)

The file starts with YAML frontmatter containing all scores in a machine-readable format:

---
type: evaluation
date: YYYY-MM-DD
agents: [vc, market-analyst, yc-founder-fit]
combined_score: 59
deal_breakers: ["Team scored 1/5"]
scores:
  vc:
    overall: 58
    dimensions:
      team: { score: 2, weight: 2.0 }
      timing: { score: 4, weight: 1.5 }
      tam: { score: 3, weight: 1.0 }
      technology: { score: 3, weight: 1.0 }
      competition: { score: 2, weight: 1.0 }
      business_model: { score: 3, weight: 1.0 }
      gtm: { score: 2, weight: 1.0 }
      traction: { score: 1, weight: 1.0 }
  market_analyst:
    overall: 66
    dimensions:
      market_size: { score: 4 }
      competitive_landscape: { score: 3 }
      timing_tailwinds: { score: 4 }
      customer_accessibility: { score: 3 }
      regulatory_risk: { score: 3 }
  yc_founder_fit:
    overall: 54
    dimensions:
      founder_market_fit: { score: 4 }
      market_size: { score: 3 }
      problem_acuteness: { score: 2 }
      competition_presence: { score: 3 }
      personal_peer_demand: { score: 2 }
      recent_possibility: { score: 4 }
      successful_proxies: { score: 3 }
      long_term_commitment: { score: 3 }
      business_scalability: { score: 2 }
      idea_space_fertility: { score: 3 }
weakest_dimension: { agent: "vc", dimension: "traction", score: 1 }
---

Below the frontmatter: the full evaluation summary table, per-agent analysis, and all per-dimension details.

Phase 5: Optional Dialogue

After presenting the report:

"These are the assessments based on your idea docs and market research. Challenge any score you disagree with, or say 'done' to finish."

If the founder challenges a score:

• Defend the assessment with evidence from the docs and web research
• If the founder provides new information that changes the analysis, revise the score
• Update the evaluation file with revised scores
• Recalculate the overall and combined scores

Session commands:

• "details [dimension]" — expand the full reasoning for a specific score
• "done" — end the evaluation session

Next Steps

After presenting the evaluation, suggest next steps based on the scores:

Evaluation complete! What's next?

→ /idea:update {idea-name}     — fix low scores by adding missing info to your docs
→ /idea:pushback {idea-name}   — stress-test the claims behind your weakest dimensions
→ /idea:forge {idea-name}      — synthesize everything if you've done multiple rounds

Choose based on the scores:

• If low scores reflect missing info (e.g., Team scored low because docs don't mention the founder's background) → suggest update
• If low scores reflect untested claims (e.g., TAM is assumed, not validated) → suggest pushback
• If scores are strong across the board → suggest forge to create the consolidated summary

Graceful Degradation

• Sparse idea folders: Distinguish "not documented" from "documented but weak." Note: "Score reflects missing documentation — [dimension] may improve if [specific doc] is fleshed out."
• Exa unavailable: Fall back to WebSearch for all research. Note once: "Exa search unavailable — using standard web search for research."
• No WebSearch available: Fall back to doc-only analysis. Note which scores lack empirical validation.
• Single agent filter: /idea:evaluate my-idea vc or market or yc produces only that agent's scores. The combined score equals the single agent's score.

If Exa tools were not available during this evaluation, append a single note at the end of the report:

Tip: Run /idea:setup to configure Exa search for richer market and competitor data.