Add production-design craft QA gate and audit current baseline

[iainherd]

Goal

Add a production-design / pixel-perfect craft QA gate to the LirCap design optimisation loop, and audit the current baseline for concrete layout, typography, illustration, breakpoint, spacing, alignment, and component-polish defects before any Pilot 001 candidate implementation proceeds.

Why this is needed

The existing loop covers brand direction, strategic fit, accessibility, performance, security, deploy smoke, and audience-agent feedback. It does not yet adequately force a production designer's eye over pixel-level craft and breakpoint behaviour.

Human review of the current baseline immediately noticed craft defects that the loop should have caught, including:

1. Navigation/menu layout with two-word labels per area feels poor and unrefined.
2. Landing-page illustration is weak design-wise, overlaps/sits awkwardly against header text, and does not resize properly.
3. Site has not been reviewed across enough common breakpoints.
4. Header text sizing/alignment varies across pages; page-top alignment shifts between routes; kerning/rhythm feels inconsistent.
5. Landing page first section has two horizontal divider treatments that do not feel intentionally related.

These are craft/production-design defects, not only brand-strategy questions. Candidate branches can have a poor direction and still be executed with craft; the process must detect both.

Required process changes

1. Add a agent:production-design-reviewer / craft QA role to the branch-scored loop.
2. Add a pixel-perfect/craft checklist covering:
   • nav layout and label rhythm;
   • hero illustration placement, scaling, overlap, and responsive behaviour;
   • common breakpoints beyond only mobile/tablet/desktop snapshots;
   • heading scale, baseline/top alignment, line-height, letter-spacing/kerning, and cross-page consistency;
   • divider/rule systems and repeated component consistency;
   • spacing grid, optical alignment, card rhythm, and vertical rhythm;
   • component state polish: hover/focus/active/reduced motion;
   • screenshot diff review before promotion.
3. Require candidate branches to pass production-design review before scoring/promotion.
4. Update templates so scorecards/run records include craft QA findings and blocking craft defects.
5. Define hard blockers for production craft regressions, including overlap, broken resizing, inconsistent typographic scale, unintentional duplicate rules/dividers, and obvious layout sloppiness.

Required current-baseline audit

Before continuing Pilot 001 implementation, perform a bounded audit of current main / pre-prod using screenshots at common breakpoints such as:

• 320
• 375/390
• 768
• 1024
• 1280
• 1366/1440
• 1728/1920 where practical

Audit at least:

• /
• /team/
• /contact/

Return concrete findings with:

• route;
• viewport;
• screenshot/evidence path;
• defect description;
• severity;
• proposed owner/profile;
• whether it blocks Pilot 001 candidate implementation or can be folded into a candidate.

Acceptance criteria

• Branch-scored loop docs/templates include a production-design / pixel-perfect craft QA gate.
• Current baseline craft audit evidence exists under the Pilot 001 loop folder or linked from the issue.
• Pilot 001 run record is updated to state candidate implementation is paused until craft defects are either built into candidate hypotheses or explicitly triaged.
• Specialist reviews confirm the new gate catches the classes of defects listed above.
• No candidate branch is promoted unless production-design review passes or any remaining craft defect is explicitly accepted with rationale.

Non-goals

• Do not perform a broad redesign in this issue.
• Do not add analytics/tracking.
• Do not replace accessibility/browser QA; production-design QA complements it.