Skip to main content

Validation Gate (Read-Only, 3 Checks)

Every preflight report passes all 3 checks before delivery. Any failure blocks delivery per HC#8. Fail-open is NOT applied since the skill is read-only.

Different from the calc-family 8-check gate because dashboard-preflight is a read-only auditing skill scoped to one dashboard, not a mutating calc author.

V1: Finding-Catalog Match​

Check: every finding in the report references one of the 8 catalog rule IDs (P-ZONE-01 through P-CONTAINER-08) OR the sentinel value "unscanned" with a rationale.

Required per finding:

  • rule_id from the 8-check catalog (P-ZONE-01, P-SHEET-02, P-FILTER-03, P-PARAM-04, P-TITLE-05, P-CANVAS-06, P-FIT-07, P-CONTAINER-08) or "unscanned".
  • severity (Blocking, Warning, Info).
  • zone_reference OR dashboard_scope: true for dashboard-level findings.
  • desktop_fix_instruction (non-empty string; concrete Desktop UI steps).
  • notes (optional free-text; required when rule_id: "unscanned").

Fail action: identify the invalid finding, re-derive the correct rule_id from the catalog, and re-run V1. If a second run still fails, refuse delivery per HC#8.

V2: Zone-Count Reconciliation​

Check: every zone identified in Phase 1 inventory appears in at least one Phase 2 finding (as a hit) or is accounted for by at least one "no finding" evaluation record.

Verification steps:

  1. Phase 1 zone count = Phase 2 accounted-for count. Mismatch is a V2 fail.
  2. dashboard_scope: true findings account for dashboard-level checks (P-CANVAS-06, dashboard-level P-TITLE-05, dashboard-level P-FILTER-03).
  3. Every worksheet-zone accounted for by at least one of P-ZONE-01, P-SHEET-02, P-TITLE-05, P-FIT-07, P-CONTAINER-08 (findings OR clean evaluations).

Fail action: identify the unaccounted zones, re-run Phase 2 checks against them, re-run V2. If still failing, refuse delivery.

V3: Handoff-Completeness​

Check: every field in dashboard_preflight_report/1 populated per skill-expertise/handoff_schema.md. Sentinel values (null, [], "unscanned") are acceptable substitutes for empty content. Blank strings and missing keys are NOT acceptable.

Required fields per schema:

  • schema_version: dashboard_preflight_report/1
  • dashboard_name, dashboard_id, workbook_name
  • zone_count, findings_count_by_severity (blocking, warning, info)
  • findings (list; may be empty if dashboard is clean)
  • unscanned_checks (list of rule IDs skipped with rationale)
  • run_metadata (timestamp, tool_failures encountered, fallback_paths used)
  • handoff_ready: true OR handoff_ready: false with a handoff_block_reason

Fail action: identify the missing fields, populate per HC#7 sentinel discipline, re-run V3. If still failing, refuse delivery.

Gate Execution Order​

  1. V1 (finding-catalog match) first, cheapest to run.
  2. V2 (zone-count reconciliation) second, validates completeness.
  3. V3 (handoff-completeness) last, depends on V1 + V2.

What V-Gate Does NOT Check​

Unlike the calc-family 8-check gate, this gate does NOT check:

  • Calc-side semantics (that's calc-family's job).
  • Runtime interaction firing (that's interaction-validator's job).
  • Content correctness of tooltips, annotations, titles (that's caption-and-tooltip-advisor, desktop-spellcheck, i18n-scanner).
  • Publish-hygiene items outside dashboard assembly (publish-readiness).

Dashboard-preflight audits assembly hygiene; it does not judge content or runtime behavior.

Refuse to Fabricate​

If V-gate cannot pass after one retry, refuse delivery per HC#8 and diagnose. Never fake a V-gate pass. Never invent a rule_id. Never omit a check without recording it as UNSCANNED.