Harden renderer verification and responsive UX

[timeleft--]

Parent

Parent PRD: #51

What to build

Harden the FAVA Rich Views renderer with end-to-end verification and responsive UX checks.

This slice should turn the individual reader surfaces into a verified milestone: realistic fixture data, failure-mode checks, Astro build verification, local server verification, desktop and mobile browser QA, and documentation of how an operator manually regenerates and serves the reader.

Acceptance criteria

• Fixture coverage includes multiple scopes, descendant scopes, decisions, observations, drafts or proposed records, superseded records, parentage, relationships, tags, multiple agents, and confidence variation.
• Tests cover duplicate thought IDs, malformed frontmatter, same-title records with different ULIDs, broken internal relationships, and route preservation.
• Tests verify the manual generation path rebuilds output from current fixture source records.
• Tests verify generated freshness metadata is present and visible.
• The Astro build passes as part of verification.
• The local CLI server path is verified against generated output.
• Browser QA covers desktop and mobile layouts with no incoherent text overlap.
• Browser QA confirms the first viewport is scannable and not a prose wall.
• Documentation explains how to manually generate and locally serve the reader.

Blocked by

• Render a descendant-aware scope dashboard #54
• Render thought detail pages with provenance and slug routes #55
• Render supersession and parentage lineage #56
• Render typed relationship navigation #57

September 7 implementation acceptance

Implemented in PR #92, merged commit043902c40e57a221dca7712021a47b9d70af785e. Independent Captain verification passed811 Python tests, required CI, complete native review, and a mobile browser flow. Synthetic Astro verification covers23 pages/251 links, four packaged templates, desktop/mobile layouts, governed current visibility, explicit historical lifecycle, full source bodies and sanitization. Source Markdown remains canonical.

This implementation slice meets its listed criteria. The five real operator sessions and outcome decision remain tracked separately in issue #54 and parent #51; synthetic verification supplies no real usage evidence.

[timeleft--]

Paused by MachineWisdomAI/WisdomLoop#35 and its five-task pilot. Minimum FAVA truth/readiness work in #72, #73, and #74 now precedes renderer and search expansion. The ready-for-agent label is removed so automated intake cannot start this work during the pilot.

Resume only if the pilot identifies human or agent retrieval as a measured bottleneck. The existing specification remains preserved as prior design context.