Arithmetic Verification & Reconciliation Checklist

Task(subagent_type="general-purpose", prompt="""
MANDATORY VERIFICATION CHECKLIST — Do not finalize the evaluation until all checks pass.

=== CHECK 1: SCORECARD ARITHMETIC ===
For each row in the IASV Scorecard:
- Verify: Score x Weight = Weighted (to 2 decimal places)
- Sum all Weighted values
- Compare to the reported Total Weighted Score
- If mismatch: FIX the Total to match the arithmetic

Show the math explicitly:
| Dimension | Score | x Weight | = Weighted | Pass? |
|-----------|-------|----------|------------|-------|
| Team      | X.X   | x 0.25   | = X.XX     |       |
| Market    | X.X   | x 0.20   | = X.XX     |       |
| Product   | X.X   | x 0.15   | = X.XX     |       |
| True Pot. | X.X   | x 0.15   | = X.XX     |       |
| Traction  | X.X   | x 0.10   | = X.XX     |       |
| Competition| X.X  | x 0.10   | = X.XX     |       |
| Terms     | X.X   | x 0.05   | = X.XX     |       |
| **TOTAL** |       |          | = **X.XX** |       |

=== CHECK 2: TPS ARITHMETIC ===
- Verify: (Defensibility x 0.35) + (Market Velocity x 0.35) + (Expansion x 0.30) = reported TPS
- If mismatch: FIX the TPS

=== CHECK 3: HEADER-BODY RECONCILIATION ===
Every value in the document header/executive summary MUST match the calculated value in the body:
- Header Scorecard score = Body scorecard total (from Check 1)
- Header TPS = Body TPS calculation (from Check 2)
- Header Council Consensus = Body MoE weighted average
- Header Confidence = Body confidence assessment
- If ANY mismatch: FIX the header to match the body

=== CHECK 4: SCALE CONSISTENCY ===
- All IASV Scorecard dimension scores: MUST be on 1-5 scale
- TPC sub-component scores (Defensibility, Market Velocity, Expansion): MUST be on 1-5 scale
- Anti-Bias MoE expert scores: MUST be on /100 scale
- Council Consensus: MUST be on /100 scale
- Flag any /10 references in scorecard scores as errors

=== CHECK 5: NO AD-HOC ADJUSTMENTS ===
- The final score is the weighted sum of the scorecard dimensions. Period.
- NO "signal premiums," "category-creation adjustments," "a16z premiums," or other post-hoc modifiers
- If a factor deserves weight, adjust the relevant dimension score instead
- Speculative "if verified" scores belong in body text only, NEVER in headers

=== CHECK 6: SIGNAL VALIDATION ===
- 3.8 - 5.0 = INVEST
- 2.8 - 3.7 = INVESTIGATE
- Below 2.8 = PASS (unless explicitly overridden with stated reason)
- If signal doesn't match score range: either fix the signal or document the override reason

=== CHECK 7: BASE SCORE IN HEADER ===
- The header reports ONLY the current base score
- Speculative scenarios ("if traction verified, score would be X.XX") belong in body text
- Never use a conditional/speculative score as the headline number

VERIFICATION RESULT:
- [ ] Check 1: Scorecard arithmetic verified
- [ ] Check 2: TPS arithmetic verified
- [ ] Check 3: Header-body reconciliation complete
- [ ] Check 4: Scale consistency confirmed
- [ ] Check 5: No ad-hoc adjustments present
- [ ] Check 6: Signal matches score range
- [ ] Check 7: Base score in header confirmed

ALL CHECKS MUST PASS before the evaluation document is written.
""")