savepoint 1.0.2 → 1.0.4

package/.claude/settings.local.json

@@ -21,7 +21,18 @@
       "Bash(go doc *)",
       "Bash(make clean *)",
       "Bash(make build-linux *)",
-      "Bash(./savepoint_test_bin --version)"
+      "Bash(./savepoint_test_bin --version)",
+      "PowerShell(go run *)",
+      "Bash(go run *)",
+      "PowerShell(Get-ChildItem -Path \"C:\\\\Users\\\\User\\\\Branding\\\\03-VIBE-LAB\\\\savepoint\" -Force | Select-Object Name, @{Name=\"Type\"; Expression={if\\($_.PSIsContainer\\){\"DIR\"}else{\"FILE\"}}} | Format-Table -AutoSize)",
+      "PowerShell(Get-ChildItem *)",
+      "Bash(Get-ChildItem -Path \"C:\\\\Users\\\\User\\\\Branding\\\\03-VIBE-LAB\\\\savepoint\\\\.savepoint\\\\releases\\\\v1.1\\\\epics\\\\E08-board-command\" -Recurse -Filter \"*.md\")",
+      "Bash(Select-Object -ExpandProperty FullName)",
+      "Bash(cmd\")",
+      "PowerShell(go build *)",
+      "PowerShell(go test *)",
+      "PowerShell(cd \"C:\\\\Users\\\\User\\\\Branding\\\\03-VIBE-LAB\\\\savepoint\"; $files = @\\(\"agent-skills\\\\savepoint-create-task\\\\SKILL.md\", \"templates\\\\project\\\\agent-skills\\\\savepoint-create-task\\\\SKILL.md\"\\); foreach \\($f in $files\\) { $bytes = [System.IO.File]::ReadAllBytes\\($f\\); Write-Host \"$f first bytes: $\\($bytes[0..9] -join ','\\) \\(CRLF: $\\($bytes[3] -eq 13\\)\\)\" })",
+      "PowerShell(cd \"C:\\\\Users\\\\User\\\\Branding\\\\03-VIBE-LAB\\\\savepoint\"; $others = Get-ChildItem agent-skills -Recurse -Filter \"SKILL.md\" | Where-Object { $_.DirectoryName -notmatch \"savepoint-create-task\" }; foreach \\($f in $others\\) { $bytes = [System.IO.File]::ReadAllBytes\\($f.FullName\\); Write-Host \"$\\($f.Name\\) dir=$\\($f.Directory.Name\\) CRLF=$\\($bytes[3] -eq 13\\)\" })"
     ]
   }
 }

package/.github/workflows/ci.yml

@@ -0,0 +1,20 @@
+name: CI
+
+on:
+  push:
+    branches: [master]
+  pull_request:
+    branches: [master]
+
+jobs:
+  ci:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-go@v5
+        with:
+          go-version: '1.26'
+          check-latest: true
+
+      - run: make ci

package/.golangci.yml

@@ -0,0 +1,11 @@
+run:
+  timeout: 5m
+
+linters:
+  enable:
+    - unused
+    - errcheck
+    - staticcheck
+    - govet
+    - ineffassign
+    - gosimple

package/.savepoint/Design.md

@@ -1,7 +1,7 @@
 ---
 type: project-design
 status: active
-last_audited: v1.1/E02-cross-platform-compatibility (2026-05-02)
+last_audited: v1.1/E15-hardening
 ---
 
 # Savepoint — System Architecture
@@ -20,11 +20,15 @@ last_audited: v1.1/E02-cross-platform-compatibility (2026-05-02)
   - Per-task incremental: <2KB.
   - Audit: 5–15KB.
   - Anything that breaks these bounds violates the wedge.
-- **Go data-reader boundary:** established in epic `E02-data-readers` (2026-05-01). `internal/data` owns Savepoint file parsing and discovery for the Go implementation: task frontmatter models, markdown YAML extraction, router state parsing, config theme defaults, release/epic/task directory listing, and boundary error sentinels.
+- **Go data-reader boundary:** established in epic `E02-data-readers` (2026-05-01). `internal/data` owns Savepoint file parsing and discovery for the Go implementation: task frontmatter models, markdown YAML extraction, router state parsing, config theme defaults, release/epic/task directory listing, task lifecycle validation/defaulting, write-time status validation, and boundary error sentinels.
 - **Template assets** live under `templates/` with helpers in `src/templates/` (epic E04).
-- **Init command** (`savepoint init`) validates, scaffolds, prints prompt, clipboard, optional install (epic E05).
-- **Board command** (`savepoint board`) reads project state, renders the Atari-Noir TUI board, supports release/epic filtering, detail overlays, task status transitions with mtime-guarded writes, release/epic-scoped router priority markers, fsnotify-based task auto-refresh (epic E06), header Next Activity display, height-aware column/detail viewport scrolling, stable focused/unfocused column border geometry (v1.1 E01), and a focusable wide-screen epic sidebar with purple epic focus, epic detail overlays, and status glyphs loaded from epic detail frontmatter (v1.1 E04).
-- **Audit pipeline** (`savepoint audit`) resolves epic, skips, quality gates, snapshots, router transition, proposal review (epic E07).
+- **Init command** (`savepoint init`) validates target directories, scaffolds rendered copies of `templates/project/`, prints the rendered magic prompt, attempts best-effort clipboard copy, and optionally runs `npm install` after scaffolding (v1.1 E07).
+- **Board command** (`savepoint board`, and bare `savepoint`) reads project state, renders the Atari-Noir TUI board when stdout is a TTY, falls back to a deterministic plain table in non-TTY mode, supports `--release`/`--epic` filtering, detail overlays, task status transitions with mtime-guarded writes, release/epic-scoped router priority markers, fsnotify-based task auto-refresh (epic E06), header Next Activity display, height-aware column/detail viewport scrolling, stable focused/unfocused column border geometry (v1.1 E01), dedicated phase-colored Next Activity line below the header, sentence-boundary checklist rendering in task details, shared status glyph mapping for task cards and the epic sidebar, a forced ANSI256 Lipgloss color profile for board startup (v1.1 E03), a focusable wide-screen epic sidebar with purple epic focus, epic detail overlays, and status glyphs loaded from epic detail frontmatter (v1.1 E04), and an epic Detail/Audit tab switch that renders user-facing audit findings from `{epic}/E##-Audit.md` (v1.1 E06).
+- **Doctor command** (`savepoint doctor`, `savepoint doctor --epic E##`) runs read-only integrity diagnostics for config, router state, release/epic/task structure, frontmatter validity, acceptance criteria presence, dependencies, duplicate task IDs, stale audit files, orphaned task IDs, and configured quality gates. It prints a human-readable report with repair suggestions and exits 0 when clean, 1 when problems are diagnosed, and 2 for internal or invocation failures.
+- **Audit remediation baseline** (v1.1 E13) centralizes frontmatter/body splitting and line-ending normalization in `internal/data`, uses typed sentinel errors for doctor repair suggestions, applies a configurable `quality_gates.gate_timeout`, removes tracked build artifacts from source control, adds `.golangci.yml`, and moves board filesystem reads/writes behind Bubble Tea command messages while preserving direct file I/O inside command helpers.
+- **Structural improvement baseline** (v1.1 E14) groups board `Model` fields into focused embedded state structs, defines consumer-side board/doctor data-access interfaces, routes doctor orphan discovery through `Discover.ListRootDirs`, renders audit-tab hidden sections via exact heading matches, improves quality-gate shell tokenization for quoted and escaped arguments, removes the separate `TaskStatus` enum in favor of `ColumnType`, and adds `internal/testutil` for shared Go test fixtures.
+- **Hardening baseline** (v1.1 E15) adds board render/layout benchmarks, data frontmatter fuzz targets, debug logging via CLI `--debug` or `SAVEPOINT_DEBUG`, abbreviation-aware task checklist sentence splitting, root test package isolation, documented audit-tab hidden-section allowlisting, repo-local CI, `make ci`, distribution SHA256 checksums, and Windows amd64/arm64 build outputs.
+- **Agent audit workflow** is skill-driven, not a CLI pipeline. At `audit-pending`, a fresh audit agent writes one epic-local `E##-Audit.md`; the user reviews its Audit tab, then asks an agent to apply the admin proposal blocks, update the visible audit findings to reflect the applied outcome, and close the epic.
 
 ## 2. Directory layout
 
@@ -37,16 +41,13 @@ last_audited: v1.1/E02-cross-platform-compatibility (2026-05-02)
     ├── visual-identity.md          ← design system; loaded conditionally for TUI work
     ├── router.md                   ← state-machine routing
     ├── config.yml                  ← theme, quality_gates, verify_strict
-    ├── audit/
-    │   └── {E##-epic}/
-    │       ├── snapshot.md
-    │       └── proposals.md
     └── releases/
         └── {release}/              ← e.g. v1, v1.1
             ├── {release}-PRD.md    ← release-scoped PRD
             └── epics/
                 └── E##-{epic-name}/
                     ├── E##-Detail.md   ← epic delta
+                    ├── E##-Audit.md    ← audit findings + admin apply proposals
                     └── tasks/
                         └── T001-slug.md
 ```
@@ -64,16 +65,18 @@ AGENTS.md at root (uppercase, cross-vendor spec). Design.md in `.savepoint/` (wo
 
 ## 4. Status model & gates
 
-Three statuses, with explicit gates:
+Three statuses, with explicit gates and ownership boundaries:
 
-| Status        | Meaning                    | Entry gate                                                      |
-| ------------- | -------------------------- | --------------------------------------------------------------- |
-| `planned`     | Ready to build             | plan section non-empty                                          |
-| `in_progress` | AI building                | all `depends_on` are `done`                                     |
-| `done`        | Complete for current scope | all implementation items checked; verification per project mode |
+| Status        | Meaning                    | Entry gate                                                      | Who may set it                         |
+| ------------- | -------------------------- | --------------------------------------------------------------- | -------------------------------------- |
+| `planned`     | Ready to build             | plan section non-empty                                          | User or planning workflow              |
+| `in_progress` | AI building                | all `depends_on` are `done`                                     | Agent when starting implementation     |
+| `done`        | Complete for current scope | all implementation items checked; verification per project mode | User only                              |
 
 - `blocked` is a **flag**, not a status — `in_progress` + `blocked: "reason"` is valid.
-- `done -> in_progress` is allowed so completed work can be reopened when follow-up work is required.
+- Agents may only advance a task into `in_progress`; they must not set `done` or retreat a task to an earlier status.
+- Only the user may set a task to `done` or retreat it from `done` to `in_progress` when follow-up work is required.
+- Router updates are explicit TUI actions: after setting a task to `in_progress`, the agent prompts the user to press `p` in the board to mark the focused task as router priority. Navigation alone must not change router task priority.
 - Verification mode: see `config.yml`.
 
 ## 5. Dependencies
@@ -82,13 +85,12 @@ Three statuses, with explicit gates:
 - `src/validation/dependencies.ts` detects duplicate task IDs, missing dependencies, and dependency cycles.
 - Cross-epic deps allowed but warned (signal that epic boundaries may be wrong).
 
-## 6. CLI surface (4 commands, no extras)
+## 6. CLI surface (3 commands, no extras)
 
 | Command                | Purpose                                                                           |
 | ---------------------- | --------------------------------------------------------------------------------- |
 | `savepoint init`       | Scaffold `.savepoint/`, print magic prompt to stdout + clipboard                  |
 | `savepoint board`      | Launch TUI; auto-falls-back to plain table on non-TTY                             |
-| `savepoint audit`      | Run audit pipeline (`--skip --reason`, `--epic`)                                  |
 | `savepoint doctor`     | Integrity check + ad-hoc quality-gate run + Layer-2 prompt for AI semantic review |
 | `--version` / `--help` | Standard global flags                                                             |
 
@@ -97,29 +99,29 @@ Three statuses, with explicit gates:
 - **Explicitly rejected:** `task new`, `epic new`, `release new`, `plan`, `next`, `status`, `task done`. All are file edits or TUI actions.
 
 **Names:** npm package `savepoint`; binary `savepoint`. No `vk` alias.
-## 7. Audit pipeline (6 steps)
+## 7. Agent audit workflow
 
 ```
-0. Quality Gates  — CLI runs configured commands. Halts on failure if block_on_failure: true.
-1. Snapshot       — CLI writes file tree (gitignore-respecting) + changed-files list. NO code contents.
-2. Diff Brief     — State flips to audit_pending: {E##-epic}. Magic prompt printed to user.
-3. Reconcile      — Agent reads epic E##-Detail.md + snapshot + scoped code. Writes one proposal bundle to
-                    .savepoint/audit/{E##-epic}/proposals.md.
-4. Review         — TUI shows side-by-side per-proposal diff. Approve / reject / edit each.
-5. Commit         — Approved proposals overwrite live files. Epic gets status: audited. Next epic unlocks.
+0. Quality Gates  — Build agent runs configured build/test gates before audit handoff.
+1. Audit Pending  — Router enters `audit-pending` for the completed epic.
+2. Reconcile      — Fresh audit agent reads router, epic detail, task files, Design.md, AGENTS.md, and scoped source/test files.
+3. Findings       — Agent writes exactly one `{epic}/E##-Audit.md`.
+4. Review         — User reviews the TUI Epic Detail Audit tab.
+5. Apply + Close  — After user approval, agent applies proposal blocks, updates the audit file's visible findings, marks the epic audited, updates `last_audited`, and advances router.
 ```
 
-- `audit_pending` is a **hard gate**: next epic's tasks cannot enter `in_progress` until prior epic is `audited`.
-- **High-divergence guard:** if a proposal changes >50% of the live file, TUI requires extra confirmation (threshold tunable in `config.yml`).
-- **Skip allowed** via `savepoint audit --skip --reason "..."`. Logged to `.savepoint/audit-log.md`. Permanent `⚠ skipped` badge in TUI.
-- **Proposal bundles** use delta-shaped edits: `Insert After`, `Replace`, or `Delete` blocks anchored to exact text.
-- **Quality review** is a section inside the proposal bundle.
-- **Snapshot availability is an audit precondition.** The router should enter `audit-pending` only after `.savepoint/audit/{release}/{E##-epic}/snapshot.md` exists.
+- `audit-pending` is a **hard gate**: next epic's tasks cannot enter `in_progress` until prior epic is `audited` or the user explicitly skips the audit.
+- `E##-Audit.md` has two user-facing sections: `## Main Findings` and `## Code Style Review`.
+- `## Main Findings` contains narrative audit findings only: AC verification, important drift, and notable risks. It must not contain file-specific replacement blocks.
+- `## Code Style Review` contains the 10 AGENTS.md code style checks.
+- `## Proposed Changes` contains admin/apply metadata using `### Target File`, `### Replace`, and `### With`. The TUI Audit tab intentionally does not render this section to avoid stale file-change details in the Epic Detail panel.
+- Apply/close must rewrite `## Main Findings` and `## Code Style Review` in the same `E##-Audit.md` so the TUI Audit tab shows resolved findings and remaining risks instead of stale pre-apply blockers. `## Proposed Changes` stays as the trace of what was applied unless the user asks otherwise.
+- There is no `savepoint audit` CLI pipeline in the active design. Audit is performed by agents using `agent-skills/savepoint-audit/SKILL.md`.
 
 Three layers:
 
 - **Layer 1 (mechanical):** user's chosen linter. Recommended: eslint+dependency-cruiser (TS), radon+pylint (Python), gocyclo+staticcheck (Go). Cross-language fallback: `lizard`. Quality gate config: see `.savepoint/config.yml`.
-- **Layer 2 (AI semantic review):** baked into the audit reconcile prompt. Outputs a quality-review section in the proposal bundle. **Advisory, not blocking.**
+- **Layer 2 (AI semantic review):** baked into the audit reconcile prompt. Outputs Main Findings and Code Style Review in the epic-local audit file. **Advisory, not blocking.**
 - **Layer 3:** `savepoint doctor` runs Layer 1 + prints Layer 2 prompt for ad-hoc use.
 
 ## 8. TUI
@@ -130,7 +132,7 @@ Acknowledged terminal limits: fonts, scanlines, glows, letter-spacing, mouse-dri
 
 **Render fallbacks:** 256-color → 16-color hard-coded → `NO_COLOR=1` monochrome with glyphs → non-TTY plain table.
 
-**Layout:** single screen with a 3-column task board (`planned`, `in_progress`, `done`), optional epic sidebar on wide terminals, centered overlays for release/epic/help/task/epic-detail views, static Atari-Noir header/footer, full-width dividers, uniform black TUI backgrounds, and navigation hints. The header can show a compact right-aligned Next Activity value from router state. Columns and detail overlays use height-aware viewport slicing with subtle above/more scroll indicators. Focused and unfocused columns preserve the same rounded-border geometry so focus changes do not shift content. On terminals at least 120 columns wide, the epic sidebar is focusable from the Planned column; it uses the purple epic accent for focused panel borders, focused epic labels, and epic detail overlays while task-column focus remains orange. Non-TTY output remains a plain table fallback.
+**Layout:** single screen with a 3-column task board (`planned`, `in_progress`, `done`), optional epic sidebar on wide terminals, centered overlays for release/epic/help/task/epic-detail views, static Atari-Noir header/footer, full-width dividers, uniform black TUI backgrounds, and navigation hints. Active router `next_action` renders as a dedicated full-width line below the header with phase-colored `PLAN`, `BUILD`, or `AUDIT` prefix styling and truncates to terminal width. Columns and detail overlays use height-aware viewport slicing with subtle above/more scroll indicators. Focused and unfocused columns preserve the same rounded-border geometry so focus changes do not shift content. Task detail implementation-plan checkboxes render once per semantic sentence, not once per hard-wrapped markdown line. On terminals at least 120 columns wide, the epic sidebar is focusable from the Planned column; it uses the purple epic accent for focused panel borders, focused epic labels, and epic detail overlays while task-column focus remains orange. Task card and epic sidebar status glyphs share `internal/board/status.go`; task cards use explicit `Task.Status` when available and retain the legacy column/stage glyph fallback when it is not. Non-TTY output remains a plain table fallback.
 
 **Visual guardrail:** the terminal board intentionally uses one black background for Background, Surface, and Surface 2. Do not restore subtly different dark panel fills; depth should come from spacing, dividers, glyphs, and focused Atari Orange borders.
 
@@ -138,11 +140,11 @@ Acknowledged terminal limits: fonts, scanlines, glows, letter-spacing, mouse-dri
 
 **Border policy:** focus must not change geometry or introduce terminal-specific broken border rendering. Use one consistent box-border family across columns, cards, and overlays. If rounded borders render as dash bars or broken segments in Warp, prefer the single-line border style already allowed by `.savepoint/visual-identity.md`; do not mix rounded and single-line borders as an ad-hoc per-component workaround.
 
-**Board persistence and refresh:** task status transitions write canonical task frontmatter through `internal/data.WriteTaskStatus` with mtime conflict checks. The board treats `Model.Root` as the `.savepoint` directory, watches `.savepoint/releases/` recursively with fsnotify, adds watches for newly-created release/epic/task directories, and reloads task plus release/epic index data plus epic status metadata after debounced file changes. Router priority markers match release + epic + task, not only the short `T###` value; completed cards render with the orange build glyph even if they previously matched router priority. Epic status glyphs are cached from each epic's `E##-Detail.md` frontmatter and shown in the wide epic sidebar only.
+**Board persistence and refresh:** task status transitions write canonical task frontmatter through `internal/data.WriteTaskStatus` with mtime conflict checks. Board update handlers dispatch filesystem reads and writes through Bubble Tea command helpers (`routerWriteMsg`, `taskWriteMsg`, `epicDetailMsg`, `auditContentMsg`, and `errorMsg`) so `Update()` remains an event/message reducer. The board treats `Model.Root` as the `.savepoint` directory, watches `.savepoint/releases/` recursively with fsnotify, adds watches for newly-created release/epic/task directories, and reloads task plus release/epic index data plus epic status metadata after debounced file changes. Router priority markers match release + epic + task, not only the short `T###` value; completed cards render with the orange build glyph even if they previously matched router priority. The `p` key explicitly writes the focused non-done task to router state as `task-building`; it does not infer `audit-pending` from task position. Epic status glyphs are cached from each epic's `E##-Detail.md` frontmatter and shown in the wide epic sidebar only.
 
 **Implementation modules:** see AGENTS.md Codebase Map.
 
-**Keybindings:** arrow/vim navigation, enter advances, backspace retreats, r/R refreshes, a/A exits toward audit review when proposals exist, q quits.
+**Keybindings:** arrow/vim navigation, enter opens focused task detail, space advances, backspace retreats, `p` marks the focused non-done task as router priority, `r`/`R` opens release selection or refreshes where supported, `?` opens help, and `q` quits or closes overlays.
 
 ## 9. Concurrency
 
@@ -177,8 +179,8 @@ All failure modes are diagnosed by `savepoint doctor`. Doctor diagnoses and prop
 - **License:** MIT.
 - **Runtime:** Go CLI binary. Source builds with `go build`; tests run with `go test ./...`.
 - **Local build:** `make build` delegates to `internal/buildtool`, builds `savepoint` or `savepoint.exe`, and injects `main.version` from `VERSION` or the latest git tag.
-- **Cross-platform builds:** `make build-all` cross-compiles linux-amd64, linux-arm64, darwin-amd64, and darwin-arm64 raw binaries into `dist/{platform}-{arch}/savepoint`.
-- **Artifacts:** `make dist` creates versioned `.tar.gz` archives in `dist/` for the Linux and Darwin targets using Go archive APIs, not shell `tar`.
+- **Cross-platform builds:** `make build-all` cross-compiles linux-amd64, linux-arm64, darwin-amd64, darwin-arm64, windows-amd64, and windows-arm64 raw binaries into `dist/{platform}-{arch}/savepoint` or `savepoint.exe` for Windows. `make ci` runs the repo-local verification sequence used by CI.
+- **Artifacts:** `make dist` creates versioned `.tar.gz` archives in `dist/` for Linux, Darwin, and Windows targets using Go archive APIs, not shell `tar`, and writes SHA256 hashes to `dist/checksums.txt`.
 - **Smoke validation:** `make smoke-test` builds the local binary and runs `--version` as a headless exit-0 check.
 - **No telemetry.** Ever.
 

package/.savepoint/{audit/v1.1/E02-cross-platform-compatibility/proposals.md → releases/v1.1/epics/E02-cross-platform-compatibility/E02-Audit.md}

@@ -1,10 +1,37 @@
-# Audit Proposals: v1.1 E02 Cross-Platform Compatibility
-
-## Target File
+---
+type: audit-findings
+audited: 2026-05-02
+---
+
+# Audit Findings: E02 Cross-Platform Compatibility
+
+## Main Findings
+
+E02 found that the cross-platform build work moved Savepoint further into its Go-based distribution path, including version handling, Linux/Darwin build targets, archive generation, and smoke-test coverage.
+
+The audit also identified that the original Makefile still retained shell-portability gaps at the time of review, including Unix-specific cleanup/archive commands and incomplete Windows packaging. Those details are retained under Proposed Changes for apply/close testing, while this visible section stays focused on the user-facing audit outcome.
+
+## Code Style Review
+
+
+- [ ] One job per file
+- [ ] One-sentence functions
+- [ ] Test branches
+- [ ] Types are documentation
+- [ ] Build, don't speculate
+- [ ] Errors at boundaries
+- [ ] One source of truth
+- [ ] Comments explain WHY
+- [ ] Content in data files
+- [ ] Small diffs
+
+## Proposed Changes
+
+### Target File
 
 `.savepoint/Design.md`
 
-## Replace
+### Replace
 
 ```md
 ## 12. Distribution & build
@@ -18,7 +45,7 @@
 - **No telemetry.** Ever.
 ```
 
-## With
+### With
 
 ```md
 ## 12. Distribution & build
@@ -34,38 +61,37 @@
 - **No telemetry.** Ever.
 ```
 
-## Target File
+---
+
+### Target File
 
 `AGENTS.md`
 
-## Replace
+### Replace
 
 ```md
 | `main.go`                            | CLI Entrypoint and root command wiring                                                               |
 ```
 
-## With
+### With
 
 ```md
 | `main.go`                            | CLI entrypoint, root command wiring, and `--version` handling via build-time version injection        |
 ```
 
-## Target File
+---
+
+### Target File
 
 `.savepoint/releases/v1.1/epics/E02-cross-platform-compatibility/E02-Detail.md`
 
-## Insert After
+### Insert After
 
 ```md
 ## Architectural notes
-
-- All Go code uses `path/filepath` — no syscalls are platform-specific
-- Bubble Tea handles terminal differences across platforms
-- No Go build tags or platform-specific files needed
-- This is purely a build-tooling and release-workflow epic
 ```
 
-## With
+### With
 
 ```md
 ## Implemented as
@@ -77,38 +103,22 @@
 - Windows zip packaging was not implemented in this epic; the final implementation covers the Linux and Darwin targets named in the epic's build-all definition of done.
 ```
 
-## Target File
+---
+
+### Target File
 
 `.savepoint/releases/v1.1/epics/E02-cross-platform-compatibility/E02-Detail.md`
 
-## Replace
+### Replace
 
 ```md
 - [x] Makefile replaced with Go-native commands (no `rm`, `cp`, `mkdir`)
 ```
 
-## With
+### With
 
 ```md
 - [ ] Makefile replaced with Go-native commands (no `rm`, `cp`, `mkdir`)
 ```
 
-## Quality Review
-
-## Must Fix Before Close
-
-- `Makefile`: `clean` still runs `rm -rf dist/`, which violates T001 acceptance criteria: "`make clean` removes the binary without using `rm -f` or other Unix-only commands" and "No shell commands in Makefile are platform-dependent."
-- `Makefile`: `build-linux` and `build-darwin` still use `mkdir -p` and inline `GOOS=... GOARCH=...` environment assignment. That is fine for Linux, macOS, Git Bash, and WSL, but it is not a shell-agnostic Makefile implementation.
-- `Makefile`: `dist` relies on `tar`, so packaging is not yet toolchain-native or Windows-shell portable.
-- `.savepoint/releases/v1.1/epics/E02-cross-platform-compatibility/tasks/T004-smoke-tests-and-artifacts.md`: the task says `go test ./...` was not run even though the task completion protocol requires the full quality-gate suite before `status: done`.
-
-## Carry Forward
-
-- Windows `.zip` artifacts were deferred in T004. Keep that as a known scope gap unless the project narrows E02 explicitly to Linux and Darwin distribution.
-- Consider moving cross-platform build and archive logic into a tiny Go helper if the project wants true shell portability while keeping `make` as the human entrypoint.
-
-## Already Fixed
-
-- `main.go` has a headless `--version` path, so smoke testing no longer needs to start the TUI.
-- Linux and Darwin raw binary targets exist for amd64 and arm64.
-- `AGENTS.md` already documents `build-all`, `dist`, and `smoke-test`.
+---

package/.savepoint/releases/v1.1/epics/E03-ui-visual-refinement/E03-Audit.md

@@ -0,0 +1,195 @@
+---
+type: audit-findings
+audited: 2026-05-02
+---
+
+# Audit Findings: E03 UI Visual Refinement
+
+## Main Findings
+
+E03 verified the visual refinement work around the board header, Next Activity line, task detail checklist rendering, shared status glyphs, and deterministic ANSI256 color profile behavior.
+
+The audit found no blocking product-code issue in the reviewed E03 scope. The main drift was documentation/process cleanup: update architecture notes to match the implemented rendering behavior, record where the color profile was actually applied, and check off the completed T006 implementation plan.
+
+## Code Style Review
+
+
+- [ ] One job per file
+- [ ] One-sentence functions
+- [ ] Test branches
+- [ ] Types are documentation
+- [ ] Build, don't speculate
+- [ ] Errors at boundaries
+- [ ] One source of truth
+- [ ] Comments explain WHY
+- [ ] Content in data files
+- [ ] Small diffs
+
+## Proposed Changes
+
+### Target File
+
+`.savepoint/Design.md`
+
+### Replace
+
+```yaml
+last_audited: v1.1/E02-cross-platform-compatibility (2026-05-02)
+```
+
+### With
+
+```yaml
+last_audited: v1.1/E03-ui-visual-refinement (2026-05-02)
+```
+
+---
+
+### Target File
+
+`.savepoint/Design.md`
+
+### Replace
+
+```markdown
+- **Board command** (`savepoint board`) reads project state, renders the Atari-Noir TUI board, supports release/epic filtering, detail overlays, task status transitions with mtime-guarded writes, release/epic-scoped router priority markers, fsnotify-based task auto-refresh (epic E06), header Next Activity display, height-aware column/detail viewport scrolling, stable focused/unfocused column border geometry (v1.1 E01), and a focusable wide-screen epic sidebar with purple epic focus, epic detail overlays, and status glyphs loaded from epic detail frontmatter (v1.1 E04).
+```
+
+### With
+
+```markdown
+- **Board command** (`savepoint board`) reads project state, renders the Atari-Noir TUI board, supports release/epic filtering, detail overlays, task status transitions with mtime-guarded writes, release/epic-scoped router priority markers, fsnotify-based task auto-refresh (epic E06), header Next Activity display, height-aware column/detail viewport scrolling, stable focused/unfocused column border geometry (v1.1 E01), dedicated phase-colored Next Activity line below the header, sentence-boundary checklist rendering in task details, shared status glyph mapping for task cards and the epic sidebar, a forced ANSI256 Lipgloss color profile for board startup (v1.1 E03), and a focusable wide-screen epic sidebar with purple epic focus, epic detail overlays, and status glyphs loaded from epic detail frontmatter (v1.1 E04).
+```
+
+---
+
+### Target File
+
+`.savepoint/Design.md`
+
+### Replace
+
+```markdown
+**Layout:** single screen with a 3-column task board (`planned`, `in_progress`, `done`), optional epic sidebar on wide terminals, centered overlays for release/epic/help/task/epic-detail views, static Atari-Noir header/footer, full-width dividers, uniform black TUI backgrounds, and navigation hints. The header can show a compact right-aligned Next Activity value from router state. Columns and detail overlays use height-aware viewport slicing with subtle above/more scroll indicators. Focused and unfocused columns preserve the same rounded-border geometry so focus changes do not shift content. On terminals at least 120 columns wide, the epic sidebar is focusable from the Planned column; it uses the purple epic accent for focused panel borders, focused epic labels, and epic detail overlays while task-column focus remains orange. Non-TTY output remains a plain table fallback.
+```
+
+### With
+
+```markdown
+**Layout:** single screen with a 3-column task board (`planned`, `in_progress`, `done`), optional epic sidebar on wide terminals, centered overlays for release/epic/help/task/epic-detail views, static Atari-Noir header/footer, full-width dividers, uniform black TUI backgrounds, and navigation hints. Active router `next_action` renders as a dedicated full-width line below the header with phase-colored `PLAN`, `BUILD`, or `AUDIT` prefix styling and truncates to terminal width. Columns and detail overlays use height-aware viewport slicing with subtle above/more scroll indicators. Focused and unfocused columns preserve the same rounded-border geometry so focus changes do not shift content. Task detail implementation-plan checkboxes render once per semantic sentence, not once per hard-wrapped markdown line. On terminals at least 120 columns wide, the epic sidebar is focusable from the Planned column; it uses the purple epic accent for focused panel borders, focused epic labels, and epic detail overlays while task-column focus remains orange. Task card and epic sidebar status glyphs share `internal/board/status.go`; task cards use explicit `Task.Status` when available and retain the legacy column/stage glyph fallback when it is not. Non-TTY output remains a plain table fallback.
+```
+
+---
+
+### Target File
+
+`AGENTS.md`
+
+### Replace
+
+```markdown
+| `internal/board/` | TUI board, overlays, epic sidebar, status glyphs |
+```
+
+## With
+
+```markdown
+| `internal/board/` | TUI board, overlays, epic sidebar, Next Activity line, detail checklist rendering, status glyphs, forced color profile |
+```
+
+---
+
+### Target File
+
+`.savepoint/releases/v1.1/epics/E03-ui-visual-refinement/E03-Detail.md`
+
+### Replace
+
+```yaml
+status: planned
+```
+
+### With
+
+```yaml
+status: audited
+```
+
+---
+
+### Target File
+
+`.savepoint/releases/v1.1/epics/E03-ui-visual-refinement/E03-Detail.md`
+
+### Replace
+
+```markdown
+| `internal/board/card.go` | Task card glyph determination (updated for shared helper) |
+| `internal/board/view_test.go` | Header, formatting, and checkbox tests |
+```
+
+### With
+
+```markdown
+| `internal/board/card.go` | Task card glyph determination (updated for shared helper) |
+| `internal/board/board.go` | Board startup; sets the Lipgloss color profile to ANSI256 before model initialization |
+| `internal/board/view_test.go` | Header, formatting, and checkbox tests |
+```
+
+---
+
+### Target File
+
+`.savepoint/releases/v1.1/epics/E03-ui-visual-refinement/E03-Detail.md`
+
+### Insert After
+
+```markdown
+| `internal/board/card.go` | Task card glyph determination (updated for shared helper) |
+| `internal/board/board.go` | Board startup; sets the Lipgloss color profile to ANSI256 before model initialization |
+| `internal/board/view_test.go` | Header, formatting, and checkbox tests |
+```
+
+### With
+
+```markdown
+## Implemented as
+
+- `internal/board/view.go` renders `next_action` as a separate line below the header through `renderNextActivityLine`, using existing footer phase styles for `PLAN`, `BUILD`, and `AUDIT`.
+- `internal/board/layout.go` accounts for the optional Next Activity line when calculating board chrome and content height.
+- `internal/data/parser.go` joins hard-wrapped checklist continuation lines before rendering so markdown wrap points do not create duplicate checklist items.
+- `internal/board/detail.go` splits checklist item text on semantic sentence boundaries and emits one `[ ]` or `[x]` marker per sentence.
+- `internal/data/task.go` adds `Task.Status` plus status constants, including `audited`, for shared board glyph rendering.
+- `internal/board/status.go` centralizes the planned, in-progress, done, and audited status glyph mapping used by task cards and the epic sidebar.
+- `internal/board/card.go` uses explicit `Task.Status` when present, while preserving the legacy column/stage glyph fallback for older task data.
+- `internal/board/epic_panel.go` delegates epic status glyph rendering to the shared status helper.
+- `internal/board/board.go` sets Lipgloss to the ANSI256 color profile at board startup. This satisfies the deterministic 256-color rendering intent, although the implementation lives at the board boundary instead of `main.go`.
+```
+
+---
+
+### Target File
+
+`.savepoint/releases/v1.1/epics/E03-ui-visual-refinement/tasks/T006-forced-256-color-profile.md`
+
+### Replace
+
+```markdown
+## Implementation Plan
+
+- [ ] Read `main.go` — understand startup flow and identify where to inject profile forcing
+- [ ] Edit `main.go` — add `lipgloss.SetColorProfile(lipgloss.Force256Color)` call before `board.Run()`
+- [ ] Run `make build && make test` to verify no regressions
+```
+
+### With
+
+```markdown
+## Implementation Plan
+
+- [x] Read board startup flow and identify where to inject profile forcing
+- [x] Edit `internal/board/board.go` — add `lipgloss.SetColorProfile(termenv.ANSI256)` before model initialization
+- [x] Run equivalent build and test gates in this Windows shell
+```
+
+---

package/.savepoint/releases/v1.1/epics/E03-ui-visual-refinement/E03-Detail.md

@@ -1,6 +1,6 @@
 ---
 type: epic-design
-status: planned
+status: audited
 ---
 
 # Epic E03: UI Visual Refinement
@@ -29,4 +29,17 @@ Polish the Atari-Noir TUI rendering across Next Activity indicator repositioning
 | `internal/data/task.go` | Task markdown parsing for checkbox placement |
 | `internal/board/status.go` | Shared status glyph mapping helper (new) |
 | `internal/board/card.go` | Task card glyph determination (updated for shared helper) |
+| `internal/board/board.go` | Board startup; sets the Lipgloss color profile to ANSI256 before model initialization |
 | `internal/board/view_test.go` | Header, formatting, and checkbox tests |
+
+## Implemented As
+
+- `internal/board/view.go` renders `next_action` as a separate line below the header through `renderNextActivityLine`, using existing footer phase styles for `PLAN`, `BUILD`, and `AUDIT`.
+- `internal/board/layout.go` accounts for the optional Next Activity line when calculating board chrome and content height.
+- `internal/data/parser.go` joins hard-wrapped checklist continuation lines before rendering so markdown wrap points do not create duplicate checklist items.
+- `internal/board/detail.go` splits checklist item text on semantic sentence boundaries and emits one `[ ]` or `[x]` marker per sentence.
+- `internal/data/task.go` adds `Task.Status` plus status constants, including `audited`, for shared board glyph rendering.
+- `internal/board/status.go` centralizes the planned, in-progress, done, and audited status glyph mapping used by task cards and the epic sidebar.
+- `internal/board/card.go` uses explicit `Task.Status` when present, while preserving the legacy column/stage glyph fallback for older task data.
+- `internal/board/epic_panel.go` delegates epic status glyph rendering to the shared status helper.
+- `internal/board/board.go` sets Lipgloss to the ANSI256 color profile at board startup. This satisfies the deterministic 256-color rendering intent, although the implementation lives at the board boundary instead of `main.go`.

package/.savepoint/releases/v1.1/epics/E03-ui-visual-refinement/tasks/T006-forced-256-color-profile.md

@@ -17,9 +17,9 @@ depends_on: []
 
 ## Implementation Plan
 
-- [ ] Read `main.go` — understand startup flow and identify where to inject profile forcing
-- [ ] Edit `main.go` — add `lipgloss.SetColorProfile(lipgloss.Force256Color)` call before `board.Run()`
-- [ ] Run `make build && make test` to verify no regressions
+- [x] Read board startup flow and identify where to inject profile forcing
+- [x] Edit `internal/board/board.go` — add `lipgloss.SetColorProfile(termenv.ANSI256)` before model initialization
+- [x] Run equivalent build and test gates in this Windows shell
 
 ## Context Log