Add a diagnostics ledger to Repository Intelligence

[szmyty]

Outcome

Add a repository-level /diagnostics/ experience to Repository Intelligence that makes bounded technical investigations resumable and auditable alongside the existing Decisions and Journey views.

This is a focused child of #27, not a second dashboard shell.

Experience

Provide:

• a filterable ledger of diagnosis records;
• lifecycle, reproduction, root-cause, remediation, and validation state shown independently;
• represented artifact and evidence freshness;
• component, environment, author, sensitivity, redaction, and related-work filters;
• a chronological investigation trail;
• competing hypothesis states and their evidence links;
• supersession and amendment history;
• deep links to the canonical Markdown record, permitted raw evidence, issues, PRs, commits, ADRs, and postmortems;
• explicit empty, unknown, stale, restricted, malformed, and not-applicable states.

The view should visually rhyme with the ADR decision ledger while preserving that diagnostics are investigations rather than accepted decisions.

Data flow

Aether contract
  → Hygiene applicability
  → Egolint validation
  → Observatory normalized projection
  → Relay static Repository Intelligence view

Do not parse or reinterpret unrestricted raw telemetry in the browser.

Safety and accessibility

• Never render secrets or restricted record bodies into public output.
• Do not display “no diagnoses” as “healthy.”
• Do not collapse partial validation into passed.
• Show why evidence is hidden without leaking sensitive counts or filenames.
• Keep every aggregate expandable to its provenance.
• Support keyboard navigation, screen readers, mobile layouts, reduced motion, and large histories.
• Build deterministically from versioned snapshots without model calls.

Acceptance criteria

• /diagnostics/ uses the shared Repository Intelligence shell and URL conventions.
• Core independent states remain visually and semantically distinct.
• Timeline and hypothesis views link to evidence without asserting truth beyond the source.
• Restricted and public projection behavior is tested.
• Empty, stale, unknown, malformed, superseded, and partial states are explicit.
• One small and one long synthetic history demonstrate usable filtering and navigation.
• Static build output is deterministic and provenance-complete.
• The feature integrates with Build the Repository Intelligence experience #27 and does not duplicate Observatory normalization.
• Reusable visual primitives are coordinated with Create reusable Repository Intelligence visualization components holon#22 when extraction is justified.

Dependencies / related

• Build the Repository Intelligence experience #27
• Render ADR history as a decision ledger at /decisions/ #30
• Define evidence-labelled diagnosis records and authoring skill aether#74
• Define repository applicability for diagnosis records hygiene#41
• Validate repository diagnosis records and derived indexes egolint#50
• Model diagnosis records in Repository Intelligence observatory#16
• Create reusable Repository Intelligence visualization components holon#22

Non-goals

• Defining diagnosis policy or schema.
• Validating diagnoses.
• Editing records in the dashboard.
• Running fixes or provider-backed analysis.
• Publishing restricted raw evidence.