artifact_id: ggsql
rubric_version: 2.8.0
formula: baseline
judges:
- openai:gpt-5.6-terra
rows:
- criterion_id: DEL-01
  category: delivery
  checker: program
  verdict: 'yes'
  points_possible: 10
  points_earned: 10.0
  evidence: ''
  agreement: null
  votes: 0
- criterion_id: DEL-02
  category: delivery
  checker: program
  verdict: 'yes'
  points_possible: 0
  points_earned: -4.0
  evidence: 1 console errors
  agreement: null
  votes: 0
- criterion_id: DEL-03
  category: delivery
  checker: program
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: ''
  agreement: null
  votes: 0
- criterion_id: COR-01
  category: correctness
  checker: program
  verdict: na
  points_possible: 0
  points_earned: 0.0
  evidence: the brief has no reference numbers
  agreement: null
  votes: 0
- criterion_id: BRF-01
  category: answers_brief
  checker: judge
  verdict: 'yes'
  points_possible: 0
  points_earned: -15.0
  evidence: “Where do things stand?” — KPI cards for Open issues, Opened / closed (28d), Awaiting triage,
    Answered within 48h (28d), Median days to close (28d), and Customer-reported open | “Is the backlog
    shrinking?” — “Open issues over time” and “Opened vs closed per week” | “Are we keeping up with triage?”
    — “Open issues by triage status and age,” “New issues answered within 48 hours,” and “Triage queue
    — oldest first” | “Where is the work?” — “Open issues by area” and “Open issues by adapter” | “How
    close are the epics?” — “Open epics by % of sub-issues closed” | “What should we work on, and who
    is on it?” — “Most-requested open issues” and “Open issues by assignee”
  agreement: 1.0
  votes: 1
- criterion_id: BRF-02
  category: answers_brief
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Compared visible requested dashboard content with visible non-requested material such as section
    headings, source/query disclosure rows, legends, and metadata.
  agreement: 1.0
  votes: 1
- criterion_id: WRK-01
  category: working
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Reviewed the visible dashboard page and its panels for a page-level crash, error state, loading
    state, or widespread missing data.
  agreement: 1.0
  votes: 1
- criterion_id: WRK-02
  category: working
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Reviewed the visible charts, KPI cards, and tables for blank panels, error messages, spinners,
    or placeholders instead of data.
  agreement: 1.0
  votes: 1
- criterion_id: HON-01
  category: honesty
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked the visible amount-encoding bar and area charts, including “Open issues over time,”
    “Opened vs closed per week,” triage-status bars, area bars, and adapter bars, for value axes truncated
    above zero.
  agreement: 1.0
  votes: 1
- criterion_id: HON-02
  category: honesty
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked titles and labels that use percentages or rates against their plotted values, including
    “New issues answered within 48 hours” and epic “Pct complete.”
  agreement: 1.0
  votes: 1
- criterion_id: HON-03
  category: honesty
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Compared repeated color-coded issue types across charts, including “Open issues over time,”
    “Open issues by area,” and “Open issues by adapter.”
  agreement: 1.0
  votes: 1
- criterion_id: HON-04
  category: honesty
  checker: judge
  verdict: na
  points_possible: 0
  points_earned: 0.0
  evidence: Checked the visible charts for two independent value axes and whether series-to-axis mapping
    was identified.
  agreement: 1.0
  votes: 1
- criterion_id: HON-05
  category: honesty
  checker: judge
  verdict: na
  points_possible: 0
  points_earned: 0.0
  evidence: Checked for side-by-side or small-multiple charts displaying the same measure on different
    unstated scales.
  agreement: 1.0
  votes: 1
- criterion_id: HON-06
  category: honesty
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked visible line charts for lines joining unordered categories. The visible line chart
    is organized by week opened.
  agreement: 1.0
  votes: 1
- criterion_id: FRM-01
  category: form
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Reviewed the forms used for trends, rankings, category comparisons, and headline metrics.
  agreement: 1.0
  votes: 1
- criterion_id: FRM-02
  category: form
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked whether two or more charts use a comparison-hiding form such as pie/donut, 3D, trend-as-table/pie,
    or ranking-as-line.
  agreement: 1.0
  votes: 1
- criterion_id: FRM-03
  category: form
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: 'Reviewed multi-series charts and their legends: issue type in the stacked area and stacked
    bars; opened versus closed weekly bars; and age bands in the triage-status bars.'
  agreement: 1.0
  votes: 1
- criterion_id: FRM-04
  category: form
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked ranked categorical comparisons, especially “Open issues by area” and “Open issues
    by adapter.” Both are displayed in descending total-open-issue order; triage statuses follow a workflow-like
    status order.
  agreement: 1.0
  votes: 1
- criterion_id: FRM-05
  category: form
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked KPI cards and charts for a lone value rendered unnecessarily as a gauge or one-bar
    chart.
  agreement: 1.0
  votes: 1
- criterion_id: FRM-07
  category: form
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Compared charts for duplicated datasets shown in the same visual form.
  agreement: 1.0
  votes: 1
- criterion_id: FRM-08
  category: form
  checker: judge
  verdict: 'yes'
  points_possible: 0
  points_earned: -8.0
  evidence: “Opened vs closed per week” — grouped bar chart; the approximately 440 opened-issues spike
    around 2026-06-01 forces the y-axis to roughly 450, flattening most other weekly opened/closed comparisons
    below about 50.
  agreement: 1.0
  votes: 1
- criterion_id: FRM-06
  category: form
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked the first and last observations of the time-series charts against the displayed as-of
    date (2026-10-14) and the chart notes, including “Complete weeks” on the weekly-flow chart.
  agreement: 1.0
  votes: 1
- criterion_id: LEG-01
  category: legibility
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Reviewed chart titles, axis ticks, legends, bar/line marks, and table headers in the full-resolution
    tiles for text-on-text or text-on-mark collisions.
  agreement: 1.0
  votes: 1
- criterion_id: LEG-02
  category: legibility
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked all visible charts for overlapping text; counted charts only where an actual text-overlap
    issue was visible.
  agreement: 1.0
  votes: 1
- criterion_id: LEG-03
  category: legibility
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Reviewed visible chart titles, labels, legends, KPI cards, and table columns within their
    respective tiles for clipping, truncation, ellipses hiding words, or overflow.
  agreement: 1.0
  votes: 1
- criterion_id: LEG-04
  category: legibility
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Used the full-resolution tiles to inspect axis labels, legends, KPI supporting text, chart
    annotations, and table text for illegibly small or faint text.
  agreement: 1.0
  votes: 1
- criterion_id: LEG-05
  category: legibility
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Reviewed the visible charts for insufficient plotting space that would make bars, lines, or
    table-like marks indistinguishable.
  agreement: 1.0
  votes: 1
- criterion_id: LBL-01
  category: labeling
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked all headline KPI tiles, chart axes, legends, subtitles, and table columns for units
    on counts, percentages, and durations.
  agreement: 1.0
  votes: 1
- criterion_id: LBL-02
  category: labeling
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked visible chart titles, axes, legends, and table headers for raw field names, code,
    or placeholders. Triage-status values such as "awaiting_triage" and "needs_repro" were treated as
    displayed category values.
  agreement: 1.0
  votes: 1
- criterion_id: LBL-03
  category: labeling
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked the page header and subtitle for scope and time coverage.
  agreement: 1.0
  votes: 1
- criterion_id: LBL-04
  category: labeling
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked the six headline tiles for a stated reporting period or page-level as-of context;
    the page states "Data as of 2026-10-14."
  agreement: 1.0
  votes: 1
- criterion_id: LBL-05
  category: labeling
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked the visible charts and tables for descriptive titles or labels, including backlog,
    weekly flow, triage, response, area, adapter, epic-progress, and requested-issues sections.
  agreement: 1.0
  votes: 1
- criterion_id: LBL-07
  category: labeling
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: 'Checked the visible percentages: "Answered within 48h (28d) 11%" and epic "Pct complete."
    The response chart states "By week opened; unanswered issues count as misses," and the epic table
    supplies "Child closed" and "Child total."'
  agreement: 1.0
  votes: 1
- criterion_id: LBL-08
  category: labeling
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked chart titles, subtitles, axes, legends, and plotted measures for mismatches, including
    the opened-versus-closed weekly bars, age-stacked triage bars, and 48-hour response-rate line.
  agreement: 1.0
  votes: 1
- criterion_id: LBL-09
  category: labeling
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Checked headline KPIs against the visible charts/tables and checked page dates and notes for
    conflicting claims.
  agreement: 1.0
  votes: 1
- criterion_id: LAY-01
  category: layout
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Space between and around the dashboard cards, charts, and tables.
  agreement: 1.0
  votes: 1
- criterion_id: LAY-02
  category: layout
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Alignment of the KPI cards, chart panels, section dividers, and tables.
  agreement: 1.0
  votes: 1
- criterion_id: LAY-04
  category: layout
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Ordering and relative prominence of KPI summaries, charts, and detailed issue tables.
  agreement: 1.0
  votes: 1
- criterion_id: LAY-05
  category: layout
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Visual weight of card borders, chart fills, backgrounds, legends, and other decoration relative
    to data.
  agreement: 1.0
  votes: 1
- criterion_id: LAY-06
  category: layout
  checker: judge
  verdict: 'no'
  points_possible: 0
  points_earned: 0.0
  evidence: Use of color across issue-type charts, age bands, response chart, progress bars, and legends.
  agreement: 1.0
  votes: 1
- criterion_id: LAY-07
  category: layout
  checker: judge
  verdict: 'yes'
  points_possible: 0
  points_earned: -4.0
  evidence: '"rendered from /home/runner/work/fusion_issue_analysis/.../fusion_issues.duckdb via ggsql
    → Vega-Lite" | Repeated "ggsql query (01_headline_kpis.sql)" / query-file disclosures beneath dashboard
    sections'
  agreement: 1.0
  votes: 1
- criterion_id: EXC-01
  category: excellence
  checker: judge
  verdict: 'no'
  points_possible: 10
  points_earned: 0.0
  evidence: Looked for a written takeaway that states a result, rather than a descriptive section or chart
    title, and checked whether a chart substantiates it.
  agreement: 1.0
  votes: 1
- criterion_id: EXC-03
  category: excellence
  checker: judge
  verdict: 'yes'
  points_possible: 3
  points_earned: 3.0
  evidence: '"OPEN ISSUES 511" with "+52 in 28 days" | "OPENED / CLOSED (28D) 123 / 71"'
  agreement: 1.0
  votes: 1
- criterion_id: EXC-04
  category: excellence
  checker: judge
  verdict: 'no'
  points_possible: 6
  points_earned: 0.0
  evidence: Reviewed charts for annotations, target/reference lines, or a deliberately highlighted spike,
    dip, period, bar, or series.
  agreement: 1.0
  votes: 1
- criterion_id: EXC-06
  category: excellence
  checker: judge
  verdict: 'no'
  points_possible: 5
  points_earned: 0.0
  evidence: Reviewed chart color treatment to determine whether neutral gray is used for most data with
    a single accent reserved for the focal subject.
  agreement: 1.0
  votes: 1
- criterion_id: EXC-07
  category: excellence
  checker: judge
  verdict: 'no'
  points_possible: 4
  points_earned: 0.0
  evidence: Reviewed the multi-series charts for direct labels on series ends or marks rather than reliance
    on legends.
  agreement: 1.0
  votes: 1
- criterion_id: EXC-08
  category: excellence
  checker: judge
  verdict: 'yes'
  points_possible: 5
  points_earned: 5.0
  evidence: '"Open engine:v2 issues in dbt-labs/dbt-core (epics excluded unless noted)" | "Open at the
    end of each week, by type" | "Complete weeks; when closed > opened the backlog shrank" | "By week
    opened; unanswered issues count as misses" | "Issues with several areas count in each" | "(none) =
    adapter-agnostic"'
  agreement: 1.0
  votes: 1
score: 0.387
category_scores:
  delivery: 0.6
  excellence: 0.242
agreement: 1.0
judge_errors: []
votes:
- criterion_id: BRF-01
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'yes'
  evidence: “Where do things stand?” — KPI cards for Open issues, Opened / closed (28d), Awaiting triage,
    Answered within 48h (28d), Median days to close (28d), and Customer-reported open | “Is the backlog
    shrinking?” — “Open issues over time” and “Opened vs closed per week” | “Are we keeping up with triage?”
    — “Open issues by triage status and age,” “New issues answered within 48 hours,” and “Triage queue
    — oldest first” | “Where is the work?” — “Open issues by area” and “Open issues by adapter” | “How
    close are the epics?” — “Open epics by % of sub-issues closed” | “What should we work on, and who
    is on it?” — “Most-requested open issues” and “Open issues by assignee”
  round: 1
- criterion_id: BRF-02
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Compared visible requested dashboard content with visible non-requested material such as section
    headings, source/query disclosure rows, legends, and metadata.
  round: 1
- criterion_id: WRK-01
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Reviewed the visible dashboard page and its panels for a page-level crash, error state, loading
    state, or widespread missing data.
  round: 1
- criterion_id: WRK-02
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Reviewed the visible charts, KPI cards, and tables for blank panels, error messages, spinners,
    or placeholders instead of data.
  round: 1
- criterion_id: HON-01
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked the visible amount-encoding bar and area charts, including “Open issues over time,”
    “Opened vs closed per week,” triage-status bars, area bars, and adapter bars, for value axes truncated
    above zero.
  round: 1
- criterion_id: HON-02
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked titles and labels that use percentages or rates against their plotted values, including
    “New issues answered within 48 hours” and epic “Pct complete.”
  round: 1
- criterion_id: HON-03
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Compared repeated color-coded issue types across charts, including “Open issues over time,”
    “Open issues by area,” and “Open issues by adapter.”
  round: 1
- criterion_id: HON-04
  judge: gpt-5.6-terra
  sample: 0
  verdict: na
  evidence: Checked the visible charts for two independent value axes and whether series-to-axis mapping
    was identified.
  round: 1
- criterion_id: HON-05
  judge: gpt-5.6-terra
  sample: 0
  verdict: na
  evidence: Checked for side-by-side or small-multiple charts displaying the same measure on different
    unstated scales.
  round: 1
- criterion_id: HON-06
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked visible line charts for lines joining unordered categories. The visible line chart
    is organized by week opened.
  round: 1
- criterion_id: FRM-01
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Reviewed the forms used for trends, rankings, category comparisons, and headline metrics.
  round: 1
- criterion_id: FRM-02
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked whether two or more charts use a comparison-hiding form such as pie/donut, 3D, trend-as-table/pie,
    or ranking-as-line.
  round: 1
- criterion_id: FRM-03
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: 'Reviewed multi-series charts and their legends: issue type in the stacked area and stacked
    bars; opened versus closed weekly bars; and age bands in the triage-status bars.'
  round: 1
- criterion_id: FRM-04
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked ranked categorical comparisons, especially “Open issues by area” and “Open issues
    by adapter.” Both are displayed in descending total-open-issue order; triage statuses follow a workflow-like
    status order.
  round: 1
- criterion_id: FRM-05
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked KPI cards and charts for a lone value rendered unnecessarily as a gauge or one-bar
    chart.
  round: 1
- criterion_id: FRM-07
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Compared charts for duplicated datasets shown in the same visual form.
  round: 1
- criterion_id: FRM-08
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'yes'
  evidence: “Opened vs closed per week” — grouped bar chart; the approximately 440 opened-issues spike
    around 2026-06-01 forces the y-axis to roughly 450, flattening most other weekly opened/closed comparisons
    below about 50.
  round: 1
- criterion_id: FRM-06
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked the first and last observations of the time-series charts against the displayed as-of
    date (2026-10-14) and the chart notes, including “Complete weeks” on the weekly-flow chart.
  round: 1
- criterion_id: LEG-01
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Reviewed chart titles, axis ticks, legends, bar/line marks, and table headers in the full-resolution
    tiles for text-on-text or text-on-mark collisions.
  round: 1
- criterion_id: LEG-02
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked all visible charts for overlapping text; counted charts only where an actual text-overlap
    issue was visible.
  round: 1
- criterion_id: LEG-03
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Reviewed visible chart titles, labels, legends, KPI cards, and table columns within their
    respective tiles for clipping, truncation, ellipses hiding words, or overflow.
  round: 1
- criterion_id: LEG-04
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Used the full-resolution tiles to inspect axis labels, legends, KPI supporting text, chart
    annotations, and table text for illegibly small or faint text.
  round: 1
- criterion_id: LEG-05
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Reviewed the visible charts for insufficient plotting space that would make bars, lines, or
    table-like marks indistinguishable.
  round: 1
- criterion_id: LBL-01
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked all headline KPI tiles, chart axes, legends, subtitles, and table columns for units
    on counts, percentages, and durations.
  round: 1
- criterion_id: LBL-02
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked visible chart titles, axes, legends, and table headers for raw field names, code,
    or placeholders. Triage-status values such as "awaiting_triage" and "needs_repro" were treated as
    displayed category values.
  round: 1
- criterion_id: LBL-03
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked the page header and subtitle for scope and time coverage.
  round: 1
- criterion_id: LBL-04
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked the six headline tiles for a stated reporting period or page-level as-of context;
    the page states "Data as of 2026-10-14."
  round: 1
- criterion_id: LBL-05
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked the visible charts and tables for descriptive titles or labels, including backlog,
    weekly flow, triage, response, area, adapter, epic-progress, and requested-issues sections.
  round: 1
- criterion_id: LBL-07
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: 'Checked the visible percentages: "Answered within 48h (28d) 11%" and epic "Pct complete."
    The response chart states "By week opened; unanswered issues count as misses," and the epic table
    supplies "Child closed" and "Child total."'
  round: 1
- criterion_id: LBL-08
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked chart titles, subtitles, axes, legends, and plotted measures for mismatches, including
    the opened-versus-closed weekly bars, age-stacked triage bars, and 48-hour response-rate line.
  round: 1
- criterion_id: LBL-09
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Checked headline KPIs against the visible charts/tables and checked page dates and notes for
    conflicting claims.
  round: 1
- criterion_id: LAY-01
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Space between and around the dashboard cards, charts, and tables.
  round: 1
- criterion_id: LAY-02
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Alignment of the KPI cards, chart panels, section dividers, and tables.
  round: 1
- criterion_id: LAY-04
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Ordering and relative prominence of KPI summaries, charts, and detailed issue tables.
  round: 1
- criterion_id: LAY-05
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Visual weight of card borders, chart fills, backgrounds, legends, and other decoration relative
    to data.
  round: 1
- criterion_id: LAY-06
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Use of color across issue-type charts, age bands, response chart, progress bars, and legends.
  round: 1
- criterion_id: LAY-07
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'yes'
  evidence: '"rendered from /home/runner/work/fusion_issue_analysis/.../fusion_issues.duckdb via ggsql
    → Vega-Lite" | Repeated "ggsql query (01_headline_kpis.sql)" / query-file disclosures beneath dashboard
    sections'
  round: 1
- criterion_id: EXC-01
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Looked for a written takeaway that states a result, rather than a descriptive section or chart
    title, and checked whether a chart substantiates it.
  round: 1
- criterion_id: EXC-03
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'yes'
  evidence: '"OPEN ISSUES 511" with "+52 in 28 days" | "OPENED / CLOSED (28D) 123 / 71"'
  round: 1
- criterion_id: EXC-04
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Reviewed charts for annotations, target/reference lines, or a deliberately highlighted spike,
    dip, period, bar, or series.
  round: 1
- criterion_id: EXC-06
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Reviewed chart color treatment to determine whether neutral gray is used for most data with
    a single accent reserved for the focal subject.
  round: 1
- criterion_id: EXC-07
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'no'
  evidence: Reviewed the multi-series charts for direct labels on series ends or marks rather than reliance
    on legends.
  round: 1
- criterion_id: EXC-08
  judge: gpt-5.6-terra
  sample: 0
  verdict: 'yes'
  evidence: '"Open engine:v2 issues in dbt-labs/dbt-core (epics excluded unless noted)" | "Open at the
    end of each week, by type" | "Complete weeks; when closed > opened the backlog shrank" | "By week
    opened; unanswered issues count as misses" | "Issues with several areas count in each" | "(none) =
    adapter-agnostic"'
  round: 1