# M Hudson Studio Quality Ratchet — Context Links

## Source Material
- /home/jgatlit/Documents/aiChemist.agency/Matt-Hudson/deliverables/2026-08-13-thu/quality-ratchet-csa-handoff.md
- /home/jgatlit/Documents/CSA/mhd-quality-ratchet/source/2026-08-13 M Hudson Studio Quality Ratchet - Handoff.md
- /home/jgatlit/Documents/CSA/mhd-quality-ratchet/source/2026-08-13 M Hudson Studio Quality Ratchet - Priority Spec.md

## Repo Ledgers — every metric in this run traces here
- /home/jgatlit/Documents/aiChemist.agency/Matt-Hudson/studio-demo/projects/_inventory/baselines.json
- /home/jgatlit/Documents/aiChemist.agency/Matt-Hudson/studio-demo/projects/_inventory/eval_records.jsonl
- /home/jgatlit/Documents/aiChemist.agency/Matt-Hudson/studio-demo/projects/_inventory/runs.jsonl
- /home/jgatlit/Documents/aiChemist.agency/Matt-Hudson/studio-demo/projects/_render/presets.py
- /home/jgatlit/Documents/aiChemist.agency/Matt-Hudson/studio-demo/projects/_render/evals.py
- /home/jgatlit/Documents/aiChemist.agency/Matt-Hudson/studio-demo/projects/_render/run_preset.py

## Vault Entities
- 300 Entities/Projects/M Hudson Design - AI Studio Engagement.md
- 300 Entities/Projects/M Hudson Design - AI Studio Engagement/2026-08-13 Priority Spec - Quality Ratchet.md

## Published Artifacts (this run)
- /home/jgatlit/Documents/CSA/mhd-quality-ratchet/output/2026-08-13 M Hudson Studio Quality Ratchet Briefing.html
- /home/jgatlit/Documents/CSA/mhd-quality-ratchet/output/2026-08-13 M Hudson Studio Quality Ratchet Input Form.html
- /home/jgatlit/Documents/CSA/mhd-quality-ratchet/output/submit.php
- /home/jgatlit/Documents/CSA/mhd-quality-ratchet/output/svg/ (7 source SVGs — 5 mermaid, 2 hand-authored charts)
- /home/jgatlit/Documents/CSA/mhd-quality-ratchet/_handoff.json

## Skill Chain
- /home/jgatlit/.claude/skills/assess-verify-and-go/SKILL.md
- /home/jgatlit/.claude/skills/project-spec-authoring/SKILL.md
- /home/jgatlit/.claude/skills/spec-to-visual-artifact/SKILL.md
- /home/jgatlit/.claude/skills/csa-pipeline/SKILL.md
- dataviz (bundled) — palette validation for the two hand-authored charts

## Git commits (engagement repo, branch vaccaro-more-walks)
- 4cbff9f — CSA handoff authored
- 189ff5c — upstream corrected to mean the INPUTS feeding the ratchet
- 1dfd099 — cross-check claim corrected against the running-floor simulation

---

## CSA Pipeline Run Summary

**Run:** M Hudson Studio Quality Ratchet
**Date:** 2026-08-13
**Pipeline:** handoff → /project-spec-authoring → /spec-to-visual-artifact → stage/publish
**QA Score:** 96/100 — PASS

### Audience split (deliberate, and inverted from the usual)
- **Briefing → Matt.** Capability narrative. No engineering asks.
- **Input Form → operator.** Five engineering decisions in Jonathan's lane. Matt decides none of them.

### Files Produced
| File | Location | Purpose |
|------|----------|---------|
| Briefing | output/2026-08-13 …Briefing.html (112 KB) | Matt-facing capability narrative |
| Input Form | output/2026-08-13 …Input Form.html (19 KB) | Operator decision capture, 5 decisions |
| submit.php | output/submit.php | Response API (object-shaped answers: choice + notes) |
| Context Links | output/2026-08-13 … - Context Links.md | This manifest |
| _handoff.json | _handoff.json | ADW conversion manifest (status: draft) |

### Diagrams Rendered (7/7)
| # | id | type | viewBox | ratio |
|---|---|---|---|---|
| 1 | input-chain | flowchart TB + 2× LR subgraph | 1084 × 386 | 2.81 |
| 2 | loop-close | flowchart LR | 1394 × 428 | 3.26 |
| 3 | gantt-runup | gantt | 960 × 316 | 3.04 |
| 4 | repo-vault | flowchart TB + `~~~` LR subgraph | 1060 × 368 | 2.88 |
| 5 | ratchet-states | stateDiagram-v2 | 609 × 678 | 0.90 |
| 6 | climb | hand-authored SVG step chart | 820 × 420 | 1.95 |
| 7 | gatebite | hand-authored SVG ranked bar | 820 × 400 | 2.05 |

No monoculture: 3 flowchart / 1 gantt / 1 state / 2 charts across 7 diagrams.

### Visual QA Results
- 3 viewports (1920×1080, 768×1024, 375×667) × 2 pages — zero horizontal overflow, zero JS errors
- 6/6 embedded SVGs render with children and non-degenerate viewBoxes
- Form interaction verified: progress counter 2/5, fill 40%, submit enables, copy-summary renders
- Live POST round-trip verified against the deployed endpoint — `{"status":"saved","decided":1}`, notes preserved

### Defects found and fixed during QA (all pre-deploy)
1. **Bar chart emitted `h --1.0`** — malformed path on the two 1-count bars (width < 4 made the closing horizontal negative-negative). Fixed with a plain rect below 9px.
2. **`direction LR` silently dropped in both subgraphs** — a single node-level edge crossing the subgraph boundary flipped the input chain from two clean rows (ratio 2.70) to unreadable columns (0.86) with overlapping edge labels. The relationship moved into node text.
3. **`<b>` rendered as literal text** — 10 occurrences in loop-close, 2 in repo-vault. Only `<br/>` is supported.
4. **Self-loop label clipped by its own node** — the premium-retry loop on the Gate diamond. Replaced with an explicit retry node, which also states the real lane policy.
5. **Two annotation collisions in the step chart** — resolved by moving explanation to HTML prose and keeping only selective direct labels.

### Correctness defect found in the source analysis
Simulating the running baseline across all 19 `premium-hero` records showed **no failing run ever
scored above the then-current floor** — so the drafted claim that a rejected take "would have raised
the baseline" was false. Corrected in the handoff (commit `1dfd099`) and the spec. The replacement
claim is stronger and ledger-backed: the two scorers caught each other in **both** directions in one
session.

### Published URLs (verified 200)
| Artifact | Public URL |
|----------|-----------|
| STEP 1 — Briefing | https://io.noboxai.com/mhd-quality-ratchet/briefing/ |
| STEP 2 — Input Form | https://io.noboxai.com/mhd-quality-ratchet/input-form/ |
| Response API | https://io.noboxai.com/mhd-quality-ratchet/submit.php |
| Export Manifest | https://io.noboxai.com/mhd-quality-ratchet/context-links.md |

### Known state
`responses.json` currently holds a **deploy smoke-test record** (respondent "Deploy Test", 1 of 5
decided). It is overwritten by the first real submission. The archive copy
`responses/response_2026-08-13_15-32-43.json` is retained as deploy evidence.