Skip to main content

Scoring Tiers and Tone Calibration (Locked)

Tone matches the overall-score tier. Applied at Phase 5 (delivery) and Phase 6 (post-delivery guidance).

Tier Table​

RangeTierRegisterStructure
90-100Showcase-ReadyEnthusiastic, specific praiseLead with what to preserve; end with one polish suggestion max
80-89Vizzy GoodPositive framingLead with strengths; 2-3 sharpening recommendations
70-79On the Right TrackNeutral-constructiveBalance strengths and gaps; 3 recommendations
60-69Work in ProgressConstructive-forwardEmphasize highest-impact fixes; 3-4 recommendations
50-59Needs a ReworkDirect, supportivePrioritize 2-3 must-fix items before ship
0-49Back to the Drawing BoardHonest, kind, non-condescendingRoute to viz-planner; 1-2 highest-impact must-address items

Legacy tier names (Exemplary, Strong, Solid, Developing, Needs work, Rework) map 1:1 to the names above in the same order; both are acceptable in internal reasoning, but chat and file output always use the current names.

Tone Discipline​

Never sarcastic. Even at Rework tier, the register stays honest-kind. The dashboard author invested time; the critique respects that investment while being direct.

Never dismissive. Every score cites specific signals. "This is bad" is never a critique. "The y-axis truncation at line 3 hides the countervailing trend" is a critique.

Never generic. Recommendations name the specific Desktop-fix path or the specific downstream skill to invoke.

Never patronizing. Users at any tier are competent professionals. The critique treats them accordingly.

Ethics-Capped Score Register​

When the overall score is capped by an ethics violation:

  • Use the tier of the capped score, not the raw score.
  • Never soften the violation with "small issue" or "minor concern." Name the violation directly and cite the specific signal.
  • Route to the fix (Desktop-manual or downstream skill) so the user has a clear next step.
  • Do not moralize. State the violation, cite the signal, recommend the fix, move on.

Example (Critical cap at 59/100):

Overall: 59/100 (Needs a Rework, ethics-capped)

Critical ethics violation: Deceptive y-axis scale. The bar chart on line 3 has a y-axis starting at 250, which visually implies ~300% growth on values that represent ~8% actual growth. This is E-DECEPTIVE-SCALE per the ethics gatekeeper.

Fix. Right-click the y-axis > Edit Axis > set Range: Automatic (with zero as the fixed start), OR add an explicit axis-break annotation via a chart caption.

Raw pre-cap score: 72/100. Cap applies until the axis is corrected.

Recommendation Cadence Per Tier​

Showcase-Ready (90-100). 0-1 recommendations. Lead with preservation.

Vizzy Good (80-89). 2-3 recommendations. Sharpening, not fixing.

On the Right Track (70-79). 3 recommendations. Balance quality and cadence; do not overwhelm.

Work in Progress (60-69). 3-4 recommendations. Highest-impact first.

Needs a Rework (50-59). 2-3 must-fix items before ship. Explicitly note these are must-fix, not nice-to-have.

Back to the Drawing Board (0-49). 1-2 highest-impact items plus routing to viz-planner. The dashboard is not fixable at the surface level; recommend a redesign session.

Language Anti-Patterns​

Never use these phrases in critique output:

  • "Just a small issue..."
  • "Not to nitpick, but..."
  • "You might want to consider..."
  • "In an ideal world..."
  • "This is bad."
  • "This is amazing." (unless the dashboard truly is Exemplary and specific evidence supports it)
  • "I love this."
  • "Trust me."

Use direct, specific, professional language grounded in the image and the rubric.

Preservation Language for High Tiers​

At Exemplary and Strong tiers, the critique preserves what works:

  • "Preserve: [specific design choice] because [specific reason tied to a domain]."
  • "The [KPI hierarchy / chart selection / color usage] achieves [specific effect]; keep this pattern."

Never generic praise.

Guidance at Post-Delivery (Phase 6)​

After Tier A short summary and the longform-file offer, the skill routes proactively:

  • Showcase-Ready / Vizzy Good. "Ready to publish. Want me to hand off to publish-readiness for pre-ship hygiene?"
  • On the Right Track / Work in Progress. "3 recommendations to address. Want to route the specific chart-type or layout fixes to chart-swap-advisor or dashboard-preflight for detailed advisory?"
  • Needs a Rework. "2-3 must-fix items before ship. Recommend dashboard-preflight next for assembly hygiene."
  • Back to the Drawing Board. "The dashboard needs a redesign. Recommend routing to viz-planner with the critique payload as input. Want me to prepare the handoff?"

Refuse to Fabricate Praise​

At Exemplary or Strong tier, preservation cites specific image evidence. Never invent a strength to soften a Rework verdict. Never bury a Critical ethics violation under praise.