docs: refresh readme screenshot and delivery evidence

CHANGELOG.md

@@ -4,8 +4,8 @@
 
 ### Added
 
-- Conservative setup integration for the ML4W network launcher, preserving custom scripts, existing backups, and write-free dry runs.
 - Process-local `dreamcoder-nmtui` launcher with terminal-native inverse selection, Fish/Zsh routing, and a Waybar default entry point; explicit Newt style overrides remain supported.
+- Conservative setup integration for the ML4W network launcher, preserving custom scripts, existing backups, and write-free dry runs.
 - PEP 621 packaging: `pyproject.toml` with setuptools build backend, `__version__` in package.
 - `dreamcoder-theme` CLI entry point (`dreamcoder-theme sync|doctor|paths`).
 - `Makefile` with targets: install, test, coverage, build, clean, lint.
@@ -24,6 +24,7 @@
 
 ### Changed
 
+- Updated the README screenshot to show Dreamcoder Workbench in Light mode with Gentle Shell and its agent workspace.
 - Refreshed the active generated theme snapshot to Dreamcoder Light while retaining the named Dark variants.
 - **Prefix restructure**: 22 directories renamed with the `Dreamcoder*` prefix (Gentleman.Dots style). Config tools (Alacritty → DreamcoderAlacritty, Kitty → DreamcoderKitty, etc.), AI configs (Pi → DreamcoderPi, Codex-*→ DreamcoderCodex*, Antigravity → DreamcoderAntigravity), and assets (themes → DreamcoderThemes, Wallpapers → DreamcoderWallpapers, profiles → DreamcoderProfiles). Infra (src, tests, installer, scripts, docs) keeps standard names.
 - Updated all internal references: Go installer, Python theme engine, scripts, tests, docs, CI workflows, .gitignore, CODEOWNERS.

README.md

@@ -17,7 +17,7 @@ Terminal-native, health-first engineering environment built on Gentleman.Dots +
 
 ## Demo
 
-<img src="dreamcoder.webp" width="600" alt="Dreamcoder Workbench" />
+<img src="dreamcoder.webp" width="600" alt="Dreamcoder Workbench in light mode with Gentle Shell and its agent workspace" />
 
 ## Índice
 

odd/tasks/dreamcoder-dark-cross-app-quality.md

@@ -0,0 +1,70 @@
+# Dreamcoder Dark cross-application quality
+
+Historical recovery record. Current delivery status is tracked in [main delivery](readme-main-delivery.md); implementation notes below are preserved.
+Objective: validate theme states and repair the separately authorized Bat/docs failures. Delivery branch: `feat/theme-quality-integration`, based on verified `origin/main`, in registered `.local-backups/worktrees/theme-quality-delivery`.
+Delivery: feature-branch-chain (user selected); separate theme, agent-color, Dark-state, Bat and documentation slices, forecast 1,000–1,400 authored lines total. Each child points to its slice; no PR/merge. Runner: pytest and existing health/link validators.
+
+## Specs
+
+S1. User goal: "perfect ahora con el theme dreamcoder dark , el objetivo ser mejor que lso mejores del mundo en theme dark". Treat this as a quality direction, not a claim that automated contrast proves a world ranking. Measure actual rendered foreground/background states; retain the existing OLED and color identity.
+
+S2. User approval: "okey te autorizo y continue". Approved scope is the preceding proposal: validate OpenCode visually when evidence is available; audit terminal/Pi first, then desktop states; clarify active/selected/focus/inactive semantics; preserve Light; provide a reproducible visual reference. Correct only measured defects. Missing runtime ownership or screenshots must be stated rather than inferred.
+
+S3. Delivery authorization: "usa buenas practicas de git y github para avanzar con los commits + push de los cambios". Use scoped commits and non-force pushes; do not publish unrelated work or create a PR/merge without authorization.
+
+S4. Additional scope: "Ampliar el alcance: investigar y corregir también Bat y documentación como una unidad aparte." Keep Bat-cache repair and the manifest/docs regression separate from theme changes.
+
+## Tasks
+
+- T1 | S1–S2 | route: delegated measurement/coverage | done | Audited 58 resolved emitted Dark text pairs; numeric floors pass. Four Starship styles reference undefined diag; exact failing regression retained for T2. Rofi/Waybar alpha composition and Herdr runtime pairing remain unproven. Commit: 2d9602235f4c8d269fa60bc6ffbfe1dd6426b681; evidence: focused exit 1 (7 pass, 1 genuine failure), related suite exit 0.
+- T2 | S1–S2 | route: delegated implementation | done | Added Dark-only diag compatibility alias; all 18 styles resolve. RED: 9 pass/1 fail; GREEN: 10 pass. Related suite, health and pre-edit Light/dusk differential pass. Commit: 2d9602235f4c8d269fa60bc6ffbfe1dd6426b681.
+- T3 | S2 | route: inline documentation | done | Added docs/configuration/dark-theme-validation.md, linked from theme-system.md. Link validator and whitespace check exit 0. No runtime change or meaningful RED for prose. Commit: 0728b11057f021adfa1e08509d80bcf7ef1a4a18.
+- T5 | S4 | route: delegated bounded repair | done | Worker RED12 failures → GREEN12 cases; related suite and health pass. Reviewed post-write ordering, isolated destinations/config and CLI error propagation. Final delivery suite under T4. Commit: fee96f5eb001a1299e7569122e589c6a628ab772.
+- T6 | S4 | route: inline bounded repair | done | Repinned ML4W fixture reproduced RED1pass/1fail. Appended stale-ref input instead of historic SHA replacement; full verifier module GREEN17 cases; inputs unchanged. Production manifest/verifier untouched. Commit: 6fe0bfb4e2127ef4bfa0d53362184c2d0c678156.
+- T7 | S3 | route: inline hook/snapshot incident repair | done | Lint/format/mypy pass; all five behavior slices passed normal commit hooks with future untracked files isolated. Commit: b5a41b99d8d676346e6edf8601b9502957b53861 (typing); 6fe0bfb4e2127ef4bfa0d53362184c2d0c678156 (final all-code hook proof).
+- T4 | S1–S4 | route: inline verification and delivery | done | All behavior/docs commits passed hooks; final checked bytes match HEAD, clean delivery worktree. Atomic non-force publication succeeded and six GitHub refs matched exact hashes. No PR/merge; original shared work retained. Commit: 0728b11057f021adfa1e08509d80bcf7ef1a4a18 (verified published delivery); final metadata receipt is this docs commit.
+
+## Log
+
+L1. Original user request, verbatim: "perfect ahora con el theme dreamcoder dark , el objetivo ser mejor que lso mejores del mundo en theme dark".
+
+L2. Subsequent user question, verbatim: "que mas planeas hacer ??". The approved proposal covered OpenCode screenshot validation; Kitty/Ghostty, Pi, tmux and Starship; Waybar, Rofi and notifications; consistent active/selected/focus/inactive semantics; comfort evaluation; and a reproducible visual reference. User approval, verbatim: "okey te autorizo y continue".
+
+L3. Existing work remains separate: earlier Light/OpenCode Dark fixes, active mode mirrors, Bat tests, ML4W docs/keybindings and unrelated task files are uncommitted in the shared tree. A new feature branch was created without moving or staging those changes. Only paths newly authored for this feature, and precise clean renderer changes if required, belong to its work unit.
+
+L4. User supplied a Dark screenshot. It confirms OLED canvas, near-black functional blocks, a legible agent label and lilac input rule, but contains no syntax-highlighted diff. Sidebar headings look weak; ownership is unconfirmed. The read-only map does not establish whether the sidebar belongs to Herdr. No screenshot-derived numeric claims or global visual acceptance are recorded.
+
+L5. Exact bounded implementation/check commands will be forwarded to workers. Existing guards determine numeric floors. Do not add speculative token cross-products, weaken thresholds, invent runtime proof, regenerate all artifacts or change Light. About 400 authored lines is a review heuristic, not a reason to remove tests or compress code. Forecast growth requires a delivery-scope decision before the next commit.
+
+L6. T1 measured 58 resolved pairs: terminal minimum 9.114, Pi 7.279 (quiet APCA minimum 54.611), tmux 5.710, resolved Starship 8.244, notification bodies 10.302. Four Starship styles (python, golang, docker_context, custom.ai_session) use undefined diag. The audit test is intentionally RED for this defect; no production GREEN claimed. Rofi/Waybar alpha surfaces require a known compositor backdrop; seven Herdr profile schemas do not prove runtime text/background ownership. New coverage: tests/test_dark_cross_app_states.py, 194 lines. Related four-module suite passes. T2 will resolve Dark references without changing Light; a Dark-only compatibility alias is a bounded option. Do not claim all applications or Light aliases were fixed.
+
+L7. T2 repaired four Dark styles through a diag alias equal to diagnostic; no colors or floors changed. All 62 audited text pairs now resolve. Focused RED 9 pass/1 fail, GREEN 10 pass; related suite and health exit 0. SHA-verified pre-edit differential preserves Light/dusk bytes. Canonical Dark Starship variant and an exact-matching active Dark artifact received only the alias. The active artifact also contains an earlier uncommitted Light-to-Dark switch: it must not be staged wholesale. Light/dusk retain the previous unresolved diag name by scope. No live shell or external configuration was changed.
+
+L8. T3 added a 48-line reachable validation reference: read-only commands, representative capture states and explicit limits for alpha compositing, inherited terminal colors, runtime ownership and visual comfort. Link validator and diff --check exit 0. Current feature-authored forecast remains below 350 lines. Delivery separates the fix/test unit from the docs unit to honor repository docs: commit policy. The dirty active Starship artifact stays uncommitted: staging its Dark alias alone into the HEAD Light palette would be incorrect.
+
+L9. Final full-suite command: PYTHONDONTWRITEBYTECODE=1 python -m pytest -p no:cacheprovider tests/ -q. Exit 1 after 54s: four failures in tests/test_bat_cache.py (including missing dreamcoder_theme.bat_cache), plus tests/test_verify_repo_sync.py::test_manifest_docs_mismatch_is_rejected. These are the same five failures recorded earlier in this session before the cross-app feature; no new failing test is reported. The known failures were not edited or retried. Native assess remains unavailable because of undeclared untracked files; its returned RDD-off plan requires author verification and no independent verifier. T4 is blocked on scope/acceptance, not marked complete. The user must choose scoped delivery with explicit known failures or authorize a separate Bat/docs repair; no commits or external deployment occurred.
+
+L10. New user authorization, verbatim: "usa buenas practicas de git y github para avanzar con los commits + push de los cambios". Additional scope, verbatim: "Ampliar el alcance: investigar y corregir también Bat y documentación como una unidad aparte." Origin fetch succeeded; verified default main. Current HEAD inherits four commits above origin/main, already published only on fix/light-tui-coherence. Publishing the current branch without an explicit base would silently include that prerequisite work. Expanded forecast exceeds 400 authored lines; ask-on-risk requires the delivery/chain decision before new commits. No force push, PR, merge, stash or destructive operation is authorized.
+
+L11. User selected "Cadena sobre una rama integradora" (feature-branch-chain). Created and registered .local-backups/worktrees/theme-quality-delivery on feat/theme-quality-integration from origin/main (53c0252), leaving the shared tree unchanged. The earlier four Light/shell commits will not be copied wholesale. Scope now includes session-owned OpenCode Light/Dark/agent-color work, Dark cross-app validation, and the separately requested Bat/docs repairs. Generated output may exceed authored line counts; prefer coherent slices rather than compressing tests to fit an arbitrary number. No commits or push yet.
+
+L12. Named session-owned theme sources/tests/docs and the explicitly authorized Bat regression test were transported into the clean worktree. The Pi renderer, shell/nmtui files, upstream docs and arbitrary active mirrors were not copied. Only the owned OpenCode integration expectation was patched into the base Pi integration test. Generated OpenCode/Codex and Starship artifacts were rebuilt locally from the generator while preserving each base file's declared mode. Current shared-tree changes remain untouched. T5 now runs in this isolated tree; T6 will make the test fixture independent of a hardcoded upstream SHA rather than reverting another feature's pins.
+
+L13. T5 worker completed in the delivery worktree: ensure_bat_cache/BatCacheError, after-all-writes integration and expanded12-case public regressions. RED exit1 (12fail), GREEN exit0 (12pass), related suite and health exit0. Initial differential comparisons exited1 on temporary paths then ownership checksums; validated normalization finally exited0 for Dark/Light, with only requested Bat status output added. Personal Bat defaults ignored; fallback destinations and missing/incomplete/failed builds covered. Bat failure is post-write, not transactional rollback. Parent reviewed module, sync and helper isolation changes; no commits/staging/push by worker.
+
+L14. T6: initial added Gentleman repin fixture remained GREEN; historic hash belongs to ML4W. Correct ML4W repin reproduced RED (exit1,1pass/1fail assert0==1). Repair appends an explicit stale ref and asserts unchanged input docs/manifest. Full verifier module GREEN exit0,17 cases. No production manifest/verifier changes. Bat public help and a short section in theme-system.md now describe cache precedence/status/failure timing. Final clean-tree links, theme health and whitespace exit0. Full suite running as job-3; no commits or push yet. Existing readline warning about cdspell is external shell-startup noise, not a test failure.
+
+
+L15. Full isolated delivery suite job-3 exited0 after71s. All collected tests passed; no failures or skips reported. Assessment and scoped commit/push are next. This proves the clean delivery worktree, not the original shared dirty tree; unrelated upstream changes remain outside this branch.
+
+
+L16. Assessment returned reviewDue=true/high_risk for the Bat subprocess boundary. Exact native status continuation exited0 with next_transition {"kind":"stop","reason_code":"rdd_disabled"}. RDD-off author-verification plan requires no independent verifier. Full suite and structural checks satisfy that plan; no native review approval is claimed. Risk: none under Task Size criteria: derived caches are rebuildable, inputs/config isolation validated, requested changes are covered, and no auth/format/concurrency/delivery tooling contract is modified.
+
+L18. Job-4 exited1 after plan commit ef4528524a6858c9ababce3e545e925939aed476. First behavior commit failed: pre-commit stashed future unstaged tracked source but left future untracked tests/modules visible, causing10 pytest failures. mypy also found the untracked agent helper's defaulted lambda type-inference error; ruff/format changed4 tests. No behavior commit or push occurred. Hooks remain enabled. Job-5 applied formatting to6 files, fixed3 lint issues and exited1 with16 findings (test-local imports, ambiguous comment punctuation and match regexes). Parent fixed those narrowly and replaced lambda with typed re.Match[str] function. Job-6 checks lint/format/types. Future owned untracked files will be parked reversibly in an ignored holding directory until their slices are introduced, with finally restoration on failure. T7 is not complete until hooks pass; T4 delivery remains pending.
+
+
+L19. Job-6 exited0: ruff, ruff-format and mypy all Passed after the owned helper/test corrections. Resuming from existing plan commit, not replaying the failed batch. Future owned untracked tests/modules and validation guide are reversibly parked until their slices are introduced; every commit keeps normal pre-commit hooks enabled. T7 remains open until the sliced snapshots pass.
+
+L20. Five behavior slices passed their ordinary hooks: surfaces ed932d789b644b30a49992ad4e7d634ee86340f4; agent colors b5a41b99d8d676346e6edf8601b9502957b53861; Dark states 2d9602235f4c8d269fa60bc6ffbfe1dd6426b681; Bat cache fee96f5eb001a1299e7569122e589c6a628ab772; docs regression 6fe0bfb4e2127ef4bfa0d53362184c2d0c678156. The final behavior hook runs the full repository suite against all final source/tests. Snapshot isolation incident is resolved with hooks intact. Post-format first slice is larger than the approximate400-line heuristic to preserve coherent renderer coverage; no tests were compressed or omitted. Docs remain a separate docs: commit. No push yet.
+
+L21. Job-7 exited0 after4m54s: every behavior commit passed ordinary hooks; final docs link checks passed, all checked source/test/doc bytes matched HEAD, delivery worktree clean, all held future files restored. Job-8 exited0: atomic non-force push succeeded and exact git ls-remote readback matched all6 integrator/behavior refs. Integrator delivery tip0728b11057f021adfa1e08509d80bcf7ef1a4a18; surfaceed932d789b644b30a49992ad4e7d634ee86340f4; agentb5a41b99d8d676346e6edf8601b9502957b53861; Dark2d9602235f4c8d269fa60bc6ffbfe1dd6426b681; Batfee96f5eb001a1299e7569122e589c6a628ab772; docs-test6fe0bfb4e2127ef4bfa0d53362184c2d0c678156. All implementation tasks are complete. This final docs-only receipt records observed code publication and supplies docs/theme-quality-validation; it contains no source changes. No force push, PR or merge; unrelated original Light/shell/ML4W work remains outside the delivered branch. Remaining visual limits are documented, not treated as automated proof.