baldart 5.0.1 → 5.2.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (58) hide show
  1. package/CHANGELOG.md +157 -0
  2. package/README.md +6 -6
  3. package/VERSION +1 -1
  4. package/framework/.claude/agents/CHANGELOG.md +90 -0
  5. package/framework/.claude/agents/REGISTRY.md +5 -5
  6. package/framework/.claude/agents/coder.md +3 -3
  7. package/framework/.claude/agents/remotion-animator-orchestrator.md +4 -3
  8. package/framework/.claude/agents/senior-researcher.md +166 -151
  9. package/framework/.claude/agents/ui-expert.md +4 -10
  10. package/framework/.claude/agents/ui-quality-critic.md +27 -11
  11. package/framework/.claude/skills/design-system-init/CHANGELOG.md +5 -0
  12. package/framework/.claude/skills/design-system-init/SKILL.md +2 -2
  13. package/framework/.claude/skills/e2e-review/CHANGELOG.md +5 -0
  14. package/framework/.claude/skills/e2e-review/SKILL.md +2 -2
  15. package/framework/.claude/skills/frontend-design/CHANGELOG.md +20 -0
  16. package/framework/.claude/skills/frontend-design/SKILL.md +39 -216
  17. package/framework/.claude/skills/gamification-design/CHANGELOG.md +5 -0
  18. package/framework/.claude/skills/gamification-design/SKILL.md +2 -2
  19. package/framework/.claude/skills/prd/CHANGELOG.md +16 -0
  20. package/framework/.claude/skills/prd/SKILL.md +1 -1
  21. package/framework/.claude/skills/prd/references/discovery-phase.md +5 -1
  22. package/framework/.claude/skills/prd/references/research-phase.md +32 -12
  23. package/framework/.claude/skills/research/CHANGELOG.md +9 -0
  24. package/framework/.claude/skills/research/SKILL.md +94 -0
  25. package/framework/.claude/skills/research/assets/report-compare.template.md +50 -0
  26. package/framework/.claude/skills/research/assets/report-decision.template.md +42 -0
  27. package/framework/.claude/skills/research/assets/report-deep.template.md +61 -0
  28. package/framework/.claude/skills/research/assets/report-regulatory.template.md +44 -0
  29. package/framework/.claude/skills/research/references/playbook.md +112 -0
  30. package/framework/.claude/skills/research/scripts/rebuild-research-index.mjs +77 -0
  31. package/framework/.claude/skills/ui-design/CHANGELOG.md +60 -1
  32. package/framework/.claude/skills/ui-design/SKILL.md +127 -75
  33. package/framework/.claude/skills/ui-design/references/anti-slop.md +106 -0
  34. package/framework/.claude/skills/ui-design/references/craft-standards.md +259 -0
  35. package/framework/.claude/skills/ui-design/references/design-brief.md +92 -0
  36. package/framework/.claude/skills/ui-design/references/design-directions.md +188 -0
  37. package/framework/.claude/skills/ui-design/references/evaluation.md +125 -165
  38. package/framework/.claude/skills/ui-design/references/generation.md +125 -92
  39. package/framework/.claude/skills/ui-design/references/inventory.md +9 -2
  40. package/framework/.claude/skills/ui-design/scripts/craft-check.mjs +248 -0
  41. package/framework/agents/component-manifest-schema.md +1 -1
  42. package/framework/agents/design-system-protocol.md +6 -6
  43. package/framework/agents/index.md +3 -0
  44. package/framework/agents/research-protocol.md +228 -0
  45. package/framework/agents/skills-mapping.md +42 -23
  46. package/framework/docs/PROJECT-CONFIGURATION.md +28 -5
  47. package/framework/templates/baldart.config.template.yml +6 -0
  48. package/framework/templates/research-index.template.md +13 -0
  49. package/framework/templates/research-sources.CHANGELOG.md +14 -0
  50. package/framework/templates/research-sources.template.md +31 -0
  51. package/package.json +1 -1
  52. package/src/commands/configure.js +27 -1
  53. package/src/commands/doctor.js +82 -0
  54. package/src/commands/update.js +48 -1
  55. package/src/utils/__tests__/classify-divergence.test.js +42 -0
  56. package/src/utils/git.js +45 -9
  57. package/src/utils/research-library.js +92 -0
  58. package/src/utils/symlinks.js +3 -0
@@ -191,38 +191,38 @@ competing methodology — it detects the task type and passes the matching
191
191
 
192
192
  ## UI/UX Skills
193
193
 
194
- ### frontend-design
195
-
196
- **When to use**:
197
-
198
- - Building UI components
199
- - Creating pages
200
- - Styling work
201
-
202
- **Triggers**:
203
-
204
- - "Create [page]"
205
- - "Build [component]"
206
- - "Design [interface]"
194
+ > **UI routing rule (since v5.2.0).** ALL design-decision work routes to
195
+ > `ui-design` (the local design studio). `frontend-design` is a RETIRED
196
+ > generator kept as a router: its broad triggers still land there, but its
197
+ > body immediately reroutes (design → `/ui-design`, approved mockup → code →
198
+ > `/ui-implement`, direct in-code UI → `ui-expert` agent per `AGENTS.md`
199
+ > § delegation). Never follow pre-v5.2.0 `frontend-design` design guidance.
200
+ > `ui-ux-pro-max` is superseded by `ui-design/references/`
201
+ > (design-directions / craft-standards / anti-slop) — never route to it.
207
202
 
208
203
  ### ui-design
209
204
 
210
205
  **When to use**:
211
206
 
212
- - Designing new pages or components (context-aware mockups)
213
- - The `/prd` skill reaches its UI design phase
214
- - Redesigning existing pages / creating design options for review
207
+ - Designing, prototyping or restyling ANY page/screen/component locally —
208
+ the internal twin of the Claude Design handoff
209
+ - The `/prd` skill reaches its UI design phase (Step 3)
210
+ - Redesigning existing pages / exploring design options or aesthetic
211
+ directions for review
215
212
 
216
213
  **Triggers**:
217
214
 
218
215
  - "/ui-design"
219
- - "Design UI" / "mockup" / "opzioni di design"
220
- - Design-heavy features, visual refinement, design-system work
216
+ - "Design UI" / "mockup" / "prototipo" / "opzioni di design" / "restyle" /
217
+ "rifai la UI" / "migliora il design" / "progetta interfaccia"
218
+ - Design-heavy features, visual refinement, greenfield landing/dashboard work
219
+
220
+ ### frontend-design (router — retired generator)
221
221
 
222
- > `ui-ux-pro-max` is a **tool used inside** the `ui-expert` agent / `ui-design`
223
- > skill (see REGISTRY.md listed as a Key Tool of `ui-expert`), not a
224
- > standalone invocable BALDART skill. Route UI work through `ui-design` /
225
- > `ui-expert`, not to `ui-ux-pro-max` directly.
222
+ **When to use**: never follow it as a workflow. If a trigger lands here
223
+ ("create/build/style [page/component/UI]"), apply its routing table:
224
+ design work `ui-design` · approved mockup `ui-implement` · in-code UI
225
+ change `ui-expert` agent · motion → `motion-design`.
226
226
 
227
227
  ### ui-implement
228
228
 
@@ -422,6 +422,23 @@ competing methodology — it detects the task type and passes the matching
422
422
  - "/capture"
423
423
  - "Salva sintesi" / "distilla conversazione" / "questa conversazione in wiki"
424
424
 
425
+ ### research
426
+
427
+ **When to use**:
428
+
429
+ - Interactive, routed research over the research library (`paths.research_dir`)
430
+ - Scopes the question, routes profile (`decision`/`deep`/`compare`/`regulatory`) +
431
+ source policy by nature, checks the library for reusable prior research
432
+ BEFORE searching, launches `senior-researcher`, files + indexes the report
433
+ - Closes the source-matrix growth loop (`SOURCE_MATRIX_CANDIDATE`)
434
+ - NOT for `/prd`'s embedded research step (prd spawns the agent directly)
435
+
436
+ **Triggers**:
437
+
438
+ - "/research"
439
+ - "Fai una ricerca" / "ricerca best practice" / "confronta X e Y"
440
+ - "È già stato ricercato?" / "cosa dice la letteratura su"
441
+
425
442
  ### graph-align
426
443
 
427
444
  **When to use**:
@@ -657,7 +674,9 @@ Is it a bug fix?
657
674
  |
658
675
  v
659
676
  Is it UI/visual work?
660
- |-- YES --> /ui-design (mockups/options) or frontend-design;
677
+ |-- YES --> design decision (new/redesign/restyle/options) /ui-design;
678
+ | approved mockup → code → /ui-implement;
679
+ | direct in-code UI change → ui-expert agent (AGENTS.md § delegation);
661
680
  | motion → /motion-design
662
681
  |
663
682
  v
@@ -94,8 +94,8 @@ Empty string `""` means the concept doesn't exist in your project. Skills gated
94
94
 
95
95
  | Key | Typical value | Used by |
96
96
  |---|---|---|
97
- | `design_system` | `docs/design-system` | ui-design, frontend-design, prd, bug, kie-ai, simplify, playwright-skill, webapp-testing, gamification-design |
98
- | `ui_guidelines` | `docs/references/ui-guidelines.md` | ui-design, frontend-design, copywriting, gamification-design |
97
+ | `design_system` | `docs/design-system` | ui-design, ui-implement, prd, bug, kie-ai, simplify, playwright-skill, webapp-testing, gamification-design |
98
+ | `ui_guidelines` | `docs/references/ui-guidelines.md` | ui-design, copywriting, gamification-design |
99
99
  | `api_index` | `docs/references/api/index.md` | doc-writing-for-rag |
100
100
  | `api_schemas` | `docs/references/api/schemas.md` | doc-writing-for-rag |
101
101
  | `api_errors` | `docs/references/errors.md` | doc-writing-for-rag |
@@ -105,6 +105,7 @@ Empty string `""` means the concept doesn't exist in your project. Skills gated
105
105
  | `backlog_dir` | `backlog` | new, prd, context-primer |
106
106
  | `adrs_dir` | `docs/decisions` | prd, prd-add |
107
107
  | `prd_dir` | `docs/prd` | prd, prd-add, ui-design |
108
+ | `research_dir` | `docs/research` | senior-researcher, research, prd (Step 2.5) — the research library (v5.1.0) |
108
109
  | `docs_dir` | `docs` | code-reviewer, senior-researcher, worktree-manager (rg-search fallback root) |
109
110
  | `references_dir` | `docs/references` | new, prd, context-primer, doc-writing-for-rag, simplify |
110
111
  | `wiki_dir` | `docs/wiki` | capture, context-primer |
@@ -129,12 +130,34 @@ e2e_review:
129
130
  | `pixel_diff_threshold` | float `0.0`–`1.0` | `0.02` | When the implementation screenshot vs mockup pixel-diff is below this fraction, the orchestrator skips the Vision call for that route and treats it as a pass. Primary latency / cost saver — most routes pass pixel-diff cleanly. Set higher to skip Vision more aggressively; set to `0.0` to always invoke Vision. |
130
131
  | `require_override_reason` | bool | `true` | When self-heal exhausts iterations and the user chooses to override the gate, a written reason is mandatory. The reason is recorded both in the batch tracker's `## Issues & Flags` (as `[E2E-OVERRIDE] <reason>`) and in `.baldart/e2e-review/<CARD-ID>/report.json`. Set `false` to allow silent override (not recommended). |
131
132
 
133
+ ### 4.2.2 `research_dir` — the research library (v5.1.0+)
134
+
135
+ Consumer-owned home of every research report produced by `senior-researcher`
136
+ (directly, via the `/research` skill, or by `/prd` Step 2.5). Layout, report
137
+ frontmatter, reuse pre-flight and the source matrix are defined in
138
+ `framework/agents/research-protocol.md`.
139
+
140
+ Lifecycle:
141
+
142
+ - **Creation** — `baldart configure` (right after writing the key; default
143
+ `docs/research` is proposed on FIRST encounter only) and `baldart doctor`
144
+ (`create-research-dir`, confirmable). Seeds `INDEX.md` (the library map) and
145
+ `SOURCES.md` (the live source matrix) from framework templates.
146
+ - **Update** — `baldart update` only backfills missing seeds into an EXISTING
147
+ directory; it never recreates a deleted library.
148
+ - **Opt-out** — set the key to `""`: respected everywhere, never re-proposed,
149
+ never resurrected.
150
+ - **Source matrix versioning** — `SOURCES.md` carries `matrix_version`
151
+ frontmatter; when the framework template ships a newer matrix, `doctor`
152
+ raises an ADVISORY (never overwrites — your `## Local additions` and manual
153
+ merges are yours).
154
+
132
155
  ### 4.3 `identity` — brand and audience facts
133
156
 
134
157
  | Key | Type | Notes |
135
158
  |---|---|---|
136
159
  | `brand_name` | string | Public product name |
137
- | `design_philosophy` | string | Free-form, one line. Examples: `"Neo-Brutalism"`, `"Glassmorphism"`, `"Minimalist"`, `"Material 3"`, `""` (no opinion). Drives aesthetic decisions in `ui-design`, `frontend-design`, `kie-ai`, `motion-design`. |
160
+ | `design_philosophy` | string | Free-form, one line. Examples: `"Neo-Brutalism"`, `"Glassmorphism"`, `"Minimalist"`, `"Material 3"`, `""` (no opinion). Drives aesthetic decisions in `ui-design`, `kie-ai`, `motion-design`. |
138
161
  | `language` | BCP-47 tag | `"en"`, `"it"`, etc. Drives copy register in `copywriting`, error-message language in `ui-design`. |
139
162
  | `audience_segments` | array of strings | Domain split, e.g. `["merchant","customer"]` or `["b2b","b2c"]`. Empty `[]` = no segmentation. Drives prompts in `copywriting`, evaluation in `ui-design`. |
140
163
 
@@ -234,7 +257,7 @@ Every flag MUST be present as `true` or `false` once `baldart configure` has run
234
257
 
235
258
  | Flag | Meaning |
236
259
  |---|---|
237
- | `has_design_system` | The project has docs rooted at `paths.design_system`. When `true`, `ui-design`, `bug`, `frontend-design`, `simplify`, `playwright-skill` etc. treat the INDEX as BLOCKING. |
260
+ | `has_design_system` | The project has docs rooted at `paths.design_system`. When `true`, `ui-design`, `ui-implement`, `bug`, `simplify`, `playwright-skill` etc. treat the INDEX as BLOCKING. |
238
261
  | `multi_tenant_theming` | The project supports per-tenant theme overrides. Activates themed-surface text/bg pairing rules in `ui-design`, `playwright-skill`, `webapp-testing`. |
239
262
  | `has_api_docs` | Canonical API docs at `paths.api_index`. Gates `doc-writing-for-rag` schemas/errors workflow. |
240
263
  | `has_backlog` | Card-driven workflow (`paths.backlog_dir` + AGENTS.md card protocol). `new` skill refuses to run without it. |
@@ -314,7 +337,7 @@ These skills had heavy hard-coding before v3.0.0; if your project has opinions i
314
337
  |---|---|
315
338
  | `ui-design` | Design philosophy mandates, theming pairing rules, charting wrappers, browser open command, evaluation thresholds |
316
339
  | `copywriting` | Brand voice pillars, audience register matrix, forbidden vocabulary, language-only rules |
317
- | `frontend-design` | Aesthetic mandates, illustration-motion contract |
340
+ | `frontend-design` | (retired to a router since v5.2.0 — port aesthetic mandates / illustration-motion contract to the `ui-design` overlay) |
318
341
  | `prd` / `prd-add` | Project terminology, mandatory review gates, design-system reads for UI PRDs |
319
342
  | `kie-ai` | Illustration character library, video-prompt motion vocabulary, forbidden aesthetics |
320
343
  | `bug` | Project debug entry points (env summary helper, cache debug switch, error-code module) |
@@ -46,6 +46,12 @@ paths:
46
46
  backlog_dir: "" # e.g. backlog
47
47
  adrs_dir: "" # e.g. docs/decisions
48
48
  prd_dir: "" # e.g. docs/prd
49
+ # Research library (since v5.1.0). Consumer-owned home of every research
50
+ # report (senior-researcher / /research skill / /prd Step 2.5), indexed for
51
+ # reuse. Seeded with INDEX.md + SOURCES.md by `configure`/`doctor`; `update`
52
+ # only backfills seeds into an existing dir. Empty = deliberate opt-out
53
+ # (never re-proposed, never resurrected). Protocol: agents/research-protocol.md.
54
+ research_dir: "" # e.g. docs/research
49
55
  docs_dir: "" # e.g. docs (umbrella docs root; rg-search fallback used by code-reviewer / senior-researcher / worktree-manager)
50
56
  references_dir: "" # e.g. docs/references
51
57
  wiki_dir: "" # e.g. docs/wiki (LLM-wiki overlay)
@@ -0,0 +1,13 @@
1
+ # Research Library — INDEX
2
+
3
+ > **This file is THE map of the research library.** Agents resolve reuse
4
+ > lookups against this index only — a report without a row here is invisible.
5
+ > Rows are **append-only**; superseded reports move to `_archive/` and their
6
+ > row is REMOVED (archive, never delete). The frontmatter of each report is
7
+ > the source of truth; this index is a regenerable view
8
+ > (`rebuild-research-index.mjs` in the `/research` skill reconciles it).
9
+ >
10
+ > Protocol: `agents/research-protocol.md SECTION=library`.
11
+
12
+ | id | title | category | profile | tags | date | valid_until | path |
13
+ |----|-------|----------|---------|------|------|-------------|------|
@@ -0,0 +1,14 @@
1
+ # Changelog — research source matrix
2
+
3
+ History of `research-sources.template.md` (`matrix_version` frontmatter).
4
+ Framework-internal (never installed into consumers — consumers hold their own
5
+ seeded `SOURCES.md` copy and merge updates via the doctor advisory). Bump
6
+ `matrix_version` in the same edit: MINOR for added sources/natures, PATCH for
7
+ demotions and note refinements.
8
+
9
+ ## 1.0.0 — 2026-07-03 (framework v5.1.0)
10
+
11
+ - Initial default matrix: 5 natures (scientific, dev, regulatory, ux, market)
12
+ × primary/secondary/avoid/notes, distilled from the v5.1.0 research-layer
13
+ design pass (OSS recon: GPT Researcher, dzhng/deep-research, STORM,
14
+ open_deep_research; house evidence-strength labels).
@@ -0,0 +1,31 @@
1
+ ---
2
+ matrix_version: 1.0.0
3
+ ---
4
+
5
+ # Research Source Matrix
6
+
7
+ > Routes each research **nature** to the right sources. Read by
8
+ > `senior-researcher` on every run (resolution order and growth loop:
9
+ > `agents/research-protocol.md SECTION=sources`). This copy is
10
+ > **consumer-owned**: BALDART seeds it once and never overwrites it. Framework
11
+ > updates to the default matrix arrive as a doctor ADVISORY (compare
12
+ > `matrix_version` against the framework template) — merging is a human
13
+ > decision. Project-specific sources go under `## Local additions`.
14
+
15
+ ## Default matrix (framework, v1.0.0)
16
+
17
+ | nature | primary | secondary | avoid | notes |
18
+ |---|---|---|---|---|
19
+ | **scientific** (algorithms, protocols, ML, formal methods) | Peer-reviewed venues (ACM DL, IEEE Xplore, USENIX, NeurIPS/ICML/ICLR), arXiv (flag as preprint), standards bodies (W3C, IETF, NIST) | Survey papers as entry points; Google Scholar citation chains; reputable lab/engineering blogs (paper-backed) | Popular-science rewrites; Medium summaries of papers; vendor "research" without artifacts | Prefer replicated results; extract method + dataset + metrics, not just conclusions |
20
+ | **dev** (frameworks, libraries, integration patterns, tooling) | Official documentation + release notes; the project's own GitHub repo (issues, discussions, RFCs); official engineering blogs | High-signal engineering blogs; conference talks; Stack Overflow accepted answers (check dates); HN threads for failure modes | SEO content farms ("Top 10 X in 2026"); AI-generated tutorial mills; outdated tutorials (> 18 months for fast-moving stacks) | Version-pin every claim (behavior changes across majors); GitHub issues reveal real gotchas docs hide |
21
+ | **regulatory** (compliance, privacy, accessibility, sector rules) | Primary legal texts (EUR-Lex, official gazettes); regulator guidance (EDPB, garanti, agencies); official standards (WCAG/W3C, PCI SSC, ISO abstracts) | Big-firm legal analyses (dated + jurisdiction-tagged); IAPP; specialized compliance blogs | Vendor compliance-marketing pages; forum legal opinions; anything without a jurisdiction + date | ALWAYS name the jurisdiction; obligations cite the primary text, practices may cite secondary |
22
+ | **ux** (interaction patterns, accessibility-in-practice, design systems) | Nielsen Norman Group; established design systems (Material, HIG, Carbon, Polaris) as pattern evidence; WCAG techniques; peer-reviewed HCI (CHI) | Baymard (e-commerce); reputable case studies with data; a11y practitioner blogs (Deque, TPGi) | Dribbble/Behance as evidence of usability (aesthetics ≠ usability); engagement-bait "UX laws" listicles | Pattern adoption across 2+ major design systems = [MODERATE] evidence of validity |
23
+ | **market** (vendor evaluation, pricing, adoption, ecosystem health) | Official pricing pages + changelogs + SLAs + status pages (dated snapshots); public financial/usage disclosures | Independent benchmarks (methodology visible); StackShare-class adoption signals; comparative engineering posts by USERS of the tool | Vendor-vs-competitor comparison pages; sponsored reviews; analyst quadrants without methodology access | Pricing claims MUST carry the retrieval date; ecosystem health = commit cadence + issue response, not stars |
24
+
25
+ ## Local additions
26
+
27
+ <!-- Project-specific sources grow here (via the /research skill's matrix loop
28
+ or by hand). Same columns. BALDART never touches this section. -->
29
+
30
+ | nature | primary | secondary | avoid | notes |
31
+ |---|---|---|---|---|
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "baldart",
3
- "version": "5.0.1",
3
+ "version": "5.2.0",
4
4
  "description": "Claude Agent Framework - Reusable framework for coordinating AI agents and humans in software projects",
5
5
  "bin": {
6
6
  "baldart": "./bin/baldart.js"
@@ -479,6 +479,7 @@ function detect(cwd = process.cwd()) {
479
479
  backlog_dir: exists('backlog') ? 'backlog' : '',
480
480
  adrs_dir: countMatches('docs/decisions', /^ADR-.*\.md$/) > 0 ? 'docs/decisions' : '',
481
481
  prd_dir: exists('docs/prd') ? 'docs/prd' : '',
482
+ research_dir: exists('docs/research') ? 'docs/research' : '',
482
483
  docs_dir: exists('docs') ? 'docs' : '',
483
484
  references_dir: exists('docs/references') ? 'docs/references' : '',
484
485
  wiki_dir: exists('docs/wiki') ? 'docs/wiki' : '',
@@ -789,7 +790,7 @@ async function interactivePrompts(merged, detected) {
789
790
  ' /design-system-init (or invoke the skill)\n' +
790
791
  'to scaffold INDEX.md + tokens-reference.md + per-component specs at\n' +
791
792
  'docs/design-system/. The registry-first protocol (ui-expert, ui-design,\n' +
792
- 'frontend-design, /design-review) needs these files to enforce coherence.'
793
+ 'ui-implement, /design-review) needs these files to enforce coherence.'
793
794
  );
794
795
  } else if (merged.features.has_design_system === false) {
795
796
  UI.info(
@@ -814,6 +815,7 @@ async function interactivePrompts(merged, detected) {
814
815
  ['backlog_dir', 'Backlog directory', () => merged.features.has_backlog],
815
816
  ['adrs_dir', 'ADR directory', () => merged.features.has_adrs],
816
817
  ['prd_dir', 'PRD directory', () => merged.features.has_prd_workflow],
818
+ ['research_dir', 'Research library directory (empty = disable)', () => true],
817
819
  ['docs_dir', 'Docs root directory (rg-search umbrella)', () => true],
818
820
  ['references_dir', 'References docs root', () => true],
819
821
  ['wiki_dir', 'Wiki overlay directory', () => merged.features.has_wiki_overlay],
@@ -1425,6 +1427,16 @@ async function configure(opts = {}) {
1425
1427
  // mergePreserving(target, source) keeps target where set, fills from source.
1426
1428
  let merged = mergePreserving(existing || {}, detected);
1427
1429
  merged = mergePreserving(merged, template);
1430
+
1431
+ // paths.research_dir (v5.1.0): default `docs/research` on FIRST encounter only
1432
+ // (key never declared in the consumer's config). An explicitly emptied key is
1433
+ // a deliberate opt-out and is respected everywhere — never re-proposed, never
1434
+ // resurrected (research-protocol.md § degrade-safe).
1435
+ const researchKeyDeclared = !!(existing && existing.paths && 'research_dir' in existing.paths);
1436
+ if (!researchKeyDeclared && merged.paths && !merged.paths.research_dir) {
1437
+ merged.paths.research_dir = 'docs/research';
1438
+ if (detected.paths) detected.paths.research_dir = 'docs/research';
1439
+ }
1428
1440
  merged.version = SCHEMA_VERSION;
1429
1441
 
1430
1442
  const dsSignals = detected.stack.design_system_signals || {};
@@ -1502,6 +1514,20 @@ async function configure(opts = {}) {
1502
1514
  UI.warning(`Agent re-merge after configure failed (${err.message}) — run \`npx baldart doctor\` to heal.`);
1503
1515
  }
1504
1516
 
1517
+ // v5.1.0 — research library: create + seed ${paths.research_dir} right after
1518
+ // the key is written. Creation ownership lives HERE (and in doctor repairs);
1519
+ // `update` only backfills seeds into an existing dir. Fail-safe: configure
1520
+ // must never die on a scaffolding hiccup — doctor heals later.
1521
+ try {
1522
+ const { ensureResearchLibrary } = require('../utils/research-library');
1523
+ const createdResearch = ensureResearchLibrary(cwd, merged, { createDir: true });
1524
+ if (createdResearch.length) {
1525
+ UI.success(`Research library initialized at ${merged.paths.research_dir}/`);
1526
+ }
1527
+ } catch (err) {
1528
+ UI.warning(`Research library setup failed (${err.message}) — run \`npx baldart doctor\` to heal.`);
1529
+ }
1530
+
1505
1531
  // Ensure overlays dir exists (user-owned)
1506
1532
  const overlaysAbs = path.join(cwd, OVERLAYS_DIR);
1507
1533
  if (!fs.existsSync(overlaysAbs)) {
@@ -656,6 +656,39 @@ async function detectState(cwd, opts = {}) {
656
656
  }
657
657
  } catch (_) { /* never block doctor on design-system probe */ }
658
658
 
659
+ // ---- Research library (since v5.1.0) --------------------------------
660
+ // Gated on paths.research_dir being SET (empty = deliberate opt-out —
661
+ // never re-proposed, per research-protocol.md § degrade-safe). Three
662
+ // probes: dir missing (repair), seeds missing (backfill), source-matrix
663
+ // version behind the framework template (ADVISORY only — merging the
664
+ // matrix is a human decision, doctor never overwrites SOURCES.md).
665
+ state.researchDirMissing = false;
666
+ state.researchSeedMissing = [];
667
+ state.researchSourcesStale = null;
668
+ try {
669
+ const researchDir = (config && !config.__malformed && config.paths && config.paths.research_dir) || '';
670
+ if (typeof researchDir === 'string' && researchDir.trim()) {
671
+ const dirAbs = path.join(cwd, researchDir.trim());
672
+ if (!fs.existsSync(dirAbs)) {
673
+ state.researchDirMissing = true;
674
+ } else {
675
+ for (const seed of ['INDEX.md', 'SOURCES.md']) {
676
+ if (!fs.existsSync(path.join(dirAbs, seed))) state.researchSeedMissing.push(seed);
677
+ }
678
+ const { readMatrixVersion } = require('../utils/research-library');
679
+ const live = readMatrixVersion(path.join(dirAbs, 'SOURCES.md'));
680
+ const tpl = readMatrixVersion(path.join(cwd, '.framework', 'framework', 'templates', 'research-sources.template.md'));
681
+ if (live && tpl && live !== tpl) {
682
+ const toNum = (v) => v.split('.').map(Number);
683
+ const [lM, lm, lp] = toNum(live); const [tM, tm, tp] = toNum(tpl);
684
+ if (tM > lM || (tM === lM && (tm > lm || (tm === lm && tp > lp)))) {
685
+ state.researchSourcesStale = { live, template: tpl };
686
+ }
687
+ }
688
+ }
689
+ }
690
+ } catch (_) { /* never block doctor on research probe */ }
691
+
659
692
  // ---- Auto-deploy allowlist presence (since v4.59.0) ----------------
660
693
  // Purely informational: git.auto_deploy bounds what `/new -auto-ship`
661
694
  // may execute as an OUTWARD action. An empty allowlist is the SAFE
@@ -1343,6 +1376,55 @@ function planActions(state) {
1343
1376
  });
1344
1377
  }
1345
1378
 
1379
+ // Research library (since v5.1.0). Creation/repair lives HERE (and in
1380
+ // configure) — update only backfills seeds into an existing dir. autoOk:
1381
+ // unlike the i18n registry ("no safe auto-create without project context"),
1382
+ // the research seeds are fully GENERIC (an empty index + the framework
1383
+ // default source matrix) — zero project context needed, so unattended
1384
+ // creation is safe. Precedent for consumer-space writes: tokens-build.
1385
+ if (state.researchDirMissing) {
1386
+ actions.push({
1387
+ key: 'create-research-dir',
1388
+ label: 'Create the research library (paths.research_dir is set but the directory is missing)',
1389
+ why: 'baldart.config.yml declares a research library but the directory does not exist on disk. Creating it (with the INDEX.md + SOURCES.md seeds and the category subdirs) restores the reuse pre-flight for senior-researcher / /research / /prd Step 2.5. If you deleted it on purpose, empty paths.research_dir instead — an empty key is a respected opt-out.',
1390
+ autoOk: true, // generic seeds, idempotent, no project context needed
1391
+ run: async () => {
1392
+ const cfg = loadConfig(state.cwd);
1393
+ const { ensureResearchLibrary } = require('../utils/research-library');
1394
+ const created = ensureResearchLibrary(state.cwd, cfg, { createDir: true });
1395
+ if (created.length) UI.success(`Research library initialized (${created.join(', ')}).`);
1396
+ else UI.info('Nothing to create — library already present.');
1397
+ },
1398
+ });
1399
+ }
1400
+ if (state.researchSeedMissing && state.researchSeedMissing.length > 0) {
1401
+ actions.push({
1402
+ key: 'backfill-research-seeds',
1403
+ label: `Backfill research library seed file(s): ${state.researchSeedMissing.join(', ')}`,
1404
+ why: `The research library directory exists but ${state.researchSeedMissing.join(' and ')} are missing (pre-v5.1 install or manual cleanup). The seeds are generic framework templates (empty index / default source matrix); existing files are never overwritten.`,
1405
+ autoOk: true, // copy-if-absent only
1406
+ run: async () => {
1407
+ const cfg = loadConfig(state.cwd);
1408
+ const { ensureResearchLibrary } = require('../utils/research-library');
1409
+ const created = ensureResearchLibrary(state.cwd, cfg, { createDir: false });
1410
+ if (created.length) UI.success(`Seeded: ${created.join(', ')}.`);
1411
+ else UI.info('Nothing to backfill.');
1412
+ },
1413
+ });
1414
+ }
1415
+ if (state.researchSourcesStale) {
1416
+ actions.push({
1417
+ key: 'research-sources-advisory',
1418
+ label: `Research source matrix behind framework default (${state.researchSourcesStale.live} < ${state.researchSourcesStale.template})`,
1419
+ why: `Your SOURCES.md carries matrix_version ${state.researchSourcesStale.live} while the framework template ships ${state.researchSourcesStale.template}. ADVISORY only — SOURCES.md is consumer-owned and doctor never overwrites it. Compare with .framework/framework/templates/research-sources.template.md (its CHANGELOG sibling lists what changed) and merge what you want, keeping your "## Local additions".`,
1420
+ autoOk: true, // advisory print-only; run() does NOT mutate anything
1421
+ run: () => {
1422
+ UI.info(`Source matrix: local ${state.researchSourcesStale.live} vs framework ${state.researchSourcesStale.template}.`);
1423
+ UI.info('Diff against .framework/framework/templates/research-sources.template.md and merge manually (your `## Local additions` section is yours).');
1424
+ },
1425
+ });
1426
+ }
1427
+
1346
1428
  // Auto-deploy allowlist (since v4.59.0). Advisory print-only — surfaced ONLY
1347
1429
  // when the allowlist is non-empty (the safe default of empty/unset opens
1348
1430
  // nothing, so it warrants no nag). Reminds the user that `/new -auto-ship`
@@ -183,6 +183,28 @@ function isBaldartManagedPath(p, extra = []) {
183
183
  return BALDART_MANAGED_PATTERNS.some((rx) => rx.test(p)) || extra.some((rx) => rx.test(p));
184
184
  }
185
185
 
186
+ // Research-library seed paths (v5.1.0). The library CONTENT is consumer-owned
187
+ // (like docs/prd/) and never enters the managed patterns; only the two seed
188
+ // files BALDART itself creates are auto-commit candidates — and only while
189
+ // UNTRACKED (see postUpdateAutoCommit), so a consumer-edited tracked
190
+ // SOURCES.md/INDEX.md is never swept into a reconcile commit.
191
+ function researchSeedPaths(cwd = process.cwd()) {
192
+ try {
193
+ const fs = require('fs');
194
+ const path = require('path');
195
+ const yaml = require('js-yaml');
196
+ const cfgPath = path.join(cwd, 'baldart.config.yml');
197
+ if (!fs.existsSync(cfgPath)) return [];
198
+ const cfg = yaml.load(fs.readFileSync(cfgPath, 'utf8'));
199
+ const dir = cfg && cfg.paths && cfg.paths.research_dir;
200
+ if (typeof dir !== 'string' || !dir.trim()) return [];
201
+ const norm = dir.trim().replace(/\\/g, '/').replace(/\/+$/, '');
202
+ return [`${norm}/INDEX.md`, `${norm}/SOURCES.md`];
203
+ } catch (_) {
204
+ return [];
205
+ }
206
+ }
207
+
186
208
  // Re-generate the root primitives (AGENTS.md + CLAUDE.md) from the versioned
187
209
  // skeletons + baldart.config.yml + overlays. Idempotent + byte-stable, so it is
188
210
  // safe to call on both the aligned-install and post-pull paths. Never throws —
@@ -252,10 +274,14 @@ async function postUpdateAutoCommit(git, newVersion, options) {
252
274
  const managed = [];
253
275
  const userOwned = [];
254
276
  const extraPatterns = i18nRegistryPatterns();
277
+ // Research seeds: managed ONLY while untracked (created by BALDART, never yet
278
+ // the consumer's). Tracked-and-modified seeds are user content — untouched.
279
+ const untracked = new Set([...status.not_added, ...status.created]);
280
+ const researchSeedNew = researchSeedPaths().filter((p) => untracked.has(p));
255
281
  for (const p of allDirty) {
256
282
  if (!p || seen.has(p)) continue;
257
283
  seen.add(p);
258
- (isBaldartManagedPath(p, extraPatterns) ? managed : userOwned).push(p);
284
+ (isBaldartManagedPath(p, extraPatterns) || researchSeedNew.includes(p) ? managed : userOwned).push(p);
259
285
  }
260
286
  if (managed.length === 0) {
261
287
  return;
@@ -986,6 +1012,17 @@ async function update(options = {}, unknownArgs = []) {
986
1012
  // --json is programmatic → force autonomous so a handwritten file is never mutated.
987
1013
  reconcileRootPrimitives(process.cwd(), alignedTools, options && options.json === true ? true : undefined);
988
1014
 
1015
+ // Research-library seed backfill (v5.1.0) — seeds only, into an EXISTING
1016
+ // dir. Never creates the dir here (creation ownership: configure/doctor;
1017
+ // update must not resurrect a deliberately deleted library). Silent + idempotent.
1018
+ try {
1019
+ const fsl = require('fs');
1020
+ const pathl = require('path');
1021
+ const yaml = require('js-yaml');
1022
+ const cfg = yaml.load(fsl.readFileSync(pathl.join(process.cwd(), 'baldart.config.yml'), 'utf8'));
1023
+ require('../utils/research-library').ensureResearchLibrary(process.cwd(), cfg, { createDir: false });
1024
+ } catch (_) { /* doctor heals */ }
1025
+
989
1026
  // Backfill hooks on the aligned/self-heal path too. Same gap class as the
990
1027
  // payload reconcile above: a consumer who framework-updated with an OLDER
991
1028
  // CLI (before a given hook shipped — e.g. the S2 Codex hooks) then aligns
@@ -1351,6 +1388,16 @@ async function update(options = {}, unknownArgs = []) {
1351
1388
  // a primitive/slot/overlay change lands. Idempotent + byte-stable.
1352
1389
  reconcileRootPrimitives(process.cwd(), enabledTools, options && options.json === true ? true : undefined);
1353
1390
 
1391
+ // Research-library seed backfill (v5.1.0) — seeds only, into an EXISTING dir
1392
+ // (creation ownership: configure/doctor; never resurrect a deleted library).
1393
+ try {
1394
+ const fsl = require('fs');
1395
+ const pathl = require('path');
1396
+ const yaml = require('js-yaml');
1397
+ const cfg = yaml.load(fsl.readFileSync(pathl.join(process.cwd(), 'baldart.config.yml'), 'utf8'));
1398
+ require('../utils/research-library').ensureResearchLibrary(process.cwd(), cfg, { createDir: false });
1399
+ } catch (_) { /* doctor heals */ }
1400
+
1354
1401
  // Routines wizard (since v2.1.0) — surfaces routines added in the new framework version
1355
1402
  try {
1356
1403
  const routinesCmd = require('./routines');
@@ -290,5 +290,47 @@ console.log('\n─── overlayCoversTouched (fs-backed) ───');
290
290
  fsMod.rmSync(tmp, { recursive: true, force: true });
291
291
  }
292
292
 
293
+ // absorbablePayloadSubset(touched) — pure. The subset whose byte-content must
294
+ // match upstream for a commit to be "absorbed". Consumer files (outside
295
+ // `.framework/`) and subtree bookkeeping (VERSION/CHANGELOG/README) are
296
+ // EXCLUDED. Fixtures reproduce the exact mayo v5.0.1→v5.1.0 divergence that
297
+ // falsely blocked the update pre-v5.1.1 (mixed FEAT commits + add-then-revert
298
+ // pairs touching VERSION/CHANGELOG).
299
+ console.log('\n─── absorbablePayloadSubset (pure) ───');
300
+ {
301
+ const cases = [
302
+ { name: 'mixed FEAT commit: framework doc + app files → only the doc survives',
303
+ touched: ['.framework/framework/agents/design-system-protocol.md',
304
+ 'backlog/FEAT-0054-03-enforcement.yml',
305
+ 'docs/references/ssot-registry.md'],
306
+ expected: ['.framework/framework/agents/design-system-protocol.md'] },
307
+ { name: 'mixed FEAT commit: framework doc + CLAUDE.md + eslint config → only the doc',
308
+ touched: ['.framework/framework/agents/design-system-protocol.md',
309
+ 'CLAUDE.md', 'eslint.responsive.config.mjs'],
310
+ expected: ['.framework/framework/agents/design-system-protocol.md'] },
311
+ { name: 'bookkeeping-only touch (VERSION/CHANGELOG/README) → empty subset',
312
+ touched: ['.framework/VERSION', '.framework/CHANGELOG.md', '.framework/README.md'],
313
+ expected: [] },
314
+ { name: 'revert commit: bookkeeping + payload scripts → only the scripts survive',
315
+ touched: ['.framework/CHANGELOG.md', '.framework/README.md', '.framework/VERSION',
316
+ '.framework/framework/.claude/skills/design-system-init/scripts/extract-ts.mjs',
317
+ '.framework/framework/agents/component-manifest-schema.md'],
318
+ expected: ['.framework/framework/.claude/skills/design-system-init/scripts/extract-ts.mjs',
319
+ '.framework/framework/agents/component-manifest-schema.md'] },
320
+ { name: 'consumer-only files → empty subset',
321
+ touched: ['src/app/page.tsx', 'README.md'], expected: [] },
322
+ { name: 'empty touched → empty subset',
323
+ touched: [], expected: [] },
324
+ { name: 'null touched → empty subset (no throw)',
325
+ touched: null, expected: [] },
326
+ ];
327
+ for (const c of cases) {
328
+ const actual = GitUtils.absorbablePayloadSubset(c.touched);
329
+ const ok = JSON.stringify(actual) === JSON.stringify(c.expected);
330
+ if (ok) { pass++; console.log(` ✓ ${c.name}`); }
331
+ else { fail++; console.log(` ✗ ${c.name} → ${JSON.stringify(actual)} (expected ${JSON.stringify(c.expected)})`); }
332
+ }
333
+ }
334
+
293
335
  console.log(`\nResult: ${pass} pass, ${fail} fail`);
294
336
  process.exit(fail > 0 ? 1 : 0);
package/src/utils/git.js CHANGED
@@ -209,10 +209,13 @@ class GitUtils {
209
209
  // - subtree-merge: `git subtree pull --squash` merge commits
210
210
  // - subtree-squash: the squash payload commits themselves
211
211
  // - chore-wrapper: CLI-generated [CHORE]/chore(baldart): commits
212
- // - absorbed: a would-be custom commit whose touched .framework/ files
213
- // are now byte-identical to upstream (FETCH_HEAD) — the edit
214
- // has been absorbed into upstream (or reset away), so a
215
- // re-sync loses nothing. Pure history, not live drift → noise.
212
+ // - absorbed: a would-be custom commit whose touched .framework/ PAYLOAD
213
+ // files are now byte-identical to upstream (FETCH_HEAD) — the
214
+ // edit has been absorbed into upstream (or reset away), so a
215
+ // re-sync loses nothing. Consumer files (outside .framework/)
216
+ // and subtree bookkeeping (VERSION/CHANGELOG/README) are not
217
+ // counted (see absorbablePayloadSubset). Pure history, not
218
+ // live drift → noise.
216
219
  // - custom-overlay-able: user edited a framework agent/skill/command
217
220
  // → should migrate to .baldart/overlays/
218
221
  // - custom-other: anything else (src/, CHANGELOG, ad-hoc fixes)
@@ -297,18 +300,51 @@ class GitUtils {
297
300
  });
298
301
  }
299
302
 
300
- // True when EVERY touched `.framework/` file's content at HEAD is already
303
+ // The subset of a commit's touched paths whose byte-content must match
304
+ // upstream for the commit to count as "absorbed". Two classes of path are
305
+ // EXCLUDED because they can never conflict in a `.framework/` subtree sync:
306
+ // 1. Anything outside `.framework/` — the consumer's OWN files. A mixed
307
+ // commit that edits both a framework doc and app code (e.g. a FEAT that
308
+ // also tweaks a protocol module) is common; the subtree merge never
309
+ // touches the app files, so they are irrelevant to whether the commit's
310
+ // framework portion is absorbed. (Pre-v5.1.1 the check bailed to
311
+ // not-absorbed on the FIRST such path → mixed commits falsely blocked
312
+ // the update even when their `.framework/` payload matched upstream.)
313
+ // 2. Subtree bookkeeping (`.framework/{VERSION,CHANGELOG.md,README.md}`) —
314
+ // framework-owned metadata the merge reconciles wholesale, which ALWAYS
315
+ // differs across versions (VERSION bumps, CHANGELOG grows). Counting it
316
+ // meant any commit touching it (e.g. an add-then-revert pair) could
317
+ // never be absorbed. It is never consumer content, so dropping it is
318
+ // safe: a hand-edit to it SHOULD be overwritten by upstream anyway.
319
+ // Pure (no fs / no git) so it stays unit-testable.
320
+ static absorbablePayloadSubset(touched) {
321
+ const prefix = `${FRAMEWORK_DIR}/`;
322
+ const META = new Set([
323
+ `${prefix}VERSION`,
324
+ `${prefix}CHANGELOG.md`,
325
+ `${prefix}README.md`,
326
+ ]);
327
+ return (touched || []).filter((p) => p.startsWith(prefix) && !META.has(p));
328
+ }
329
+
330
+ // True when EVERY `.framework/` PAYLOAD file touched by the commit is already
301
331
  // byte-identical to the upstream (FETCH_HEAD) blob — i.e. the commit's edit
302
332
  // has been fully absorbed into upstream (or reset away) and a re-sync would
303
333
  // lose nothing. Compares blob SHAs (cheap, exact). The subtree prefix
304
334
  // `.framework/` is stripped to reach the upstream path (FETCH_HEAD has no
305
- // `.framework/` prefix). Conservative: any file outside `.framework/`, missing
306
- // upstream, or whose SHA differs not absorbed (real drift, keep flagging).
335
+ // `.framework/` prefix). The comparison runs over `absorbablePayloadSubset`
336
+ // ONLY (consumer files + subtree bookkeeping excluded see there). A commit
337
+ // whose payload subset is EMPTY (it touched `.framework/` only via bookkeeping
338
+ // and/or only touched consumer files) carries no subtree drift → absorbed.
339
+ // Conservative on the payload itself: a file missing at HEAD (deleted) or
340
+ // missing upstream (net-new), or whose SHA differs → not absorbed (real
341
+ // drift, keep flagging).
307
342
  async isAbsorbedAgainstUpstream(touched) {
308
343
  if (!touched || !touched.length) return false;
309
344
  const prefix = `${FRAMEWORK_DIR}/`;
310
- for (const p of touched) {
311
- if (!p.startsWith(prefix)) return false;
345
+ const payload = GitUtils.absorbablePayloadSubset(touched);
346
+ if (!payload.length) return true;
347
+ for (const p of payload) {
312
348
  const upstreamPath = p.slice(prefix.length);
313
349
  let headSha;
314
350
  let upSha;