- sync-manifest.json: controls which files sync, to which channel
- sync-from-manifest.js: reads manifest, dedup by hash, replaces old versions
- GitHub Action: auto-syncs on push to main (agents, skills, workflows)
- Channels: stable (all), beta (opt-in), internal (not distributed)
Push to main → Action runs → Design Space updated → all agents get new version.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- All bmad-manifest.json files bumped to 0.4.1
- Design Space manifest: work-orders replaced with update-status
- Added "For Agents Without Specific Instructions" section
to Design Space SKILL.md — any agent entering the system
gets a clear guide on how messaging works
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Everything is a message. Work orders use message_type: "work-order".
Check returns ALL messages — signal strength highlights, never hides.
New update-status action. Legacy task actions preserved for compat.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Transforms the Platform Requirements steps from a technology shopping
list into a complete build specification that a coding agent can
execute autonomously.
New steps:
- 28a: Data Model — complete database schema (every table, column, type,
default, constraint, relationship, index, release phase tags)
- 28b: State Machines — every stateful entity gets states, transitions,
triggers, per-phase action differences, feature gate pattern
- 28c: Business Logic, Events & API — calculations as pseudocode,
event system (scheduled/webhook/user), notification rules, complete
API surface with request/response shapes, auth spec, AI integration
Updated steps:
- 28: Tech Stack now includes architecture pattern and API style
- 32: Finalization now includes the Litmus Test — "Can a coding agent
build the entire platform from this document alone?"
Inspired by RightsTrak session where Saga produced a tech selection
document and Mårten had to push four times to get a build spec.
The 200-line tech list became a 700-line actionable specification.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Design Space (Agent Space) is now a self-contained BMad skill in WDS:
- Semantic knowledge search, capture, cross-agent messaging
- Session-scoped presence (saga-2567), work orders, online roster
- Embedded HTTP credentials — works everywhere without MCP
- Will be placed in _bmad/wds/ by installer when selected
Also bumps version to 0.3.5 including Saga/Freya skill conversion.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
New skill structure:
- src/skills/saga/SKILL.md — full agent with Product Brief + Trigger Map
- src/skills/freya/SKILL.md — full agent with UX Scenarios + UX Design
- bmad-manifest.json for each with capability menus
- references/ with all agent guides copied into skill directories
Skills are self-contained — no SKILL.md → workflow.md routing.
Follows BMad skill format: 2-part description, activation modes,
progressive disclosure, references loaded on demand.
Old activation files (saga.activation.md, freya.activation.md) kept
for now — can be removed after installer is updated.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When users provide existing materials (PRD, brief, research), Saga now:
1. Reads and extracts information into categories (vision, users, etc.)
2. Presents findings ONE category at a time for user confirmation
3. Identifies gaps after all confirmed categories
4. Plans which steps need full conversation vs quick confirmation
5. Skips exploratory questions for already-confirmed topics
Updated 13 files:
- working-with-existing-materials.md: complete rewrite with new protocol
- step-01-init.md: Material Analysis Phase trigger after welcome
- Steps 02-11: "Check Material Analysis Status" gate before exploration
Fixes the "black hole" problem where Saga acknowledged documents
but then asked questions whose answers were in those documents.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Idunn (PM agent) has been deprecated. Freya now handles design delivery,
product evolution, and agentic development. Platform requirements moved
to Saga under Phase 1. Deleted 9 Idunn-owned files, updated 33 files
across src/ and docs/.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move to private WDS working folder (martens-documents/WDS/Archive/):
- Eira visual designer gem
- Agent dialog system (templates, steps, workflows)
- docs/examples (WDS-Presentation, v6-conversion archives)
These are working/historic files, not part of the published module.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace Agent Dialog Gate with Design Log entry point in all 29 activity
workflows (phases 2-8). Replace "Update agent dialog" with "Update design log"
in 90+ step files across phases 5, 7, and 8. Update data files, guides,
presentations, and templates to reference design log instead of agent dialogs.
Only the deprecated _agent-dialogs/ archive folder retains original references.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove Agent Dialog Gate from all 9 activity workflows and the main
workflow.md initialization. Replace with a single 00-design-log.md file
containing Backlog, Current, Design Loop Status, and Log sections.
- Rewrite workflow.md adaptive dashboard to read from design log only
- Create 00-design-log.template.md, delete old 00-progress.template.md
- Update installer.js: agent-experiences folder replaces agent-dialogs
- Rename 00-progress.md → 00-design-log.md across all phases (1-5)
- Update step files to use agent experiences instead of agent dialogs
- 30 files changed, net reduction of ~140 lines of ceremony
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Path fixes across 14 files:
- D-UX-Design/ → C-UX-Scenarios/ (page specs) or D-Design-System/ (tokens)
- 4-scenarios/ → C-UX-Scenarios/ (all step files and page creation flows)
New spacing & typography step in [P] Specify:
- step-08-spacing-typography.md defines section spacing objects and heading tokens
- Old step-08 renumbered to step-09, references updated to steps 1-8
[M] Design System extraction rules:
- Objects extract on 2nd use, spacing extracts immediately on 1st use
- Component extraction check before designing 2nd+ page
[S] Suggest step-15 alignment:
- Old [A]/[S]/[M] menu replaced with two-option transitions
- Matches the pattern used by [C] Conceptualize and workflow-suggest.md
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Page spec template: rename -gap to -space, use {page}-{v|h}-{type}-{size} naming
- Design system template: objects extract on 2nd use, spacing extracts immediately
- Pattern examples updated with Above/Below/Why table format
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Spacing gets the same treatment as typography — named tokens, a baseline
(space-md), and optical adjustments using token math. But the real shift:
spacing between objects is now a first-class object with its own ID in
page specs, organized in a Patterns section in the design system.
Key additions:
- Spacing Scale: 9 tokens (space-3xs to space-3xl), bring-your-own supported
- Type Scale: 9 tokens (text-3xs to text-3xl), semantic/visual separation
- Optical adjustments: token math (space-lg - space-3xs) with annotations
- Spacing objects in page specs: ↕ `id` — token, including space-zero
- Patterns section: organized by spacing value, context added on violation
- Agent reflection behavior: observe and reflect, don't interrogate
- Design loop: pattern recognition emerges from real design decisions
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Design system template: spacing scale (9 tokens, space-3xs to space-3xl)
and type scale (9 tokens, text-3xs to text-3xl) with bring-your-own option.
Semantic heading level separated from visual text sizing — H1 can look
different on every page. Page spec template: spacing, typography, and
per-section padding/gap using token names. Design loop guide: 9 steps
(added APPROVE — user exports PNG as approval gate), spacing discipline,
export=approval pattern. Wireframe flow: agent draws, user reviews, user
exports PNG to confirm, then spec syncs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace static activity menu with adaptive dashboard that reads agent dialogs
(session plans) and per-page design log to suggest where to continue. Add
explicit design log reporting points with exact file paths and formats to all
Phase 4 and 5 workflow files. Replace all "Return to activity menu" dead ends
with two-option transitions. Rename Conceptualize → Discuss. Implements
learnings from Källa Fordonservice project.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The VTC (Value Trigger Chain) was a separate concept that extracted
specific nodes from the Trigger Map into a YAML file. This caused
agents to run trigger mapping during Phase 1 (Product Brief), forcing
users to repeat the work in Phase 2. Observed in Monday workshop.
Architectural fix:
- Phase 1 is casual business discovery, NOT trigger mapping
- Phase 2 is where the Trigger Map gets built (unchanged)
- When scenarios/content need strategic context, they reference
the Trigger Map directly — no separate VTC concept needed
Removed (19 files deleted):
- 1-project-brief/vtc-workshop/ (14 step files + workflow)
- step-04-create-vtc.md and step-11a-create-vtc.md from Phase 1
- step-03c-create-vtc.md from Phase 0 alignment
- vtc-template.yaml
Renamed:
- step-03-build-vtcs.md → step-03-build-strategic-context.md
- step-02-vtc-consistency.md → step-02-trigger-map-consistency.md
- step-01-load-vtc-context.md → step-01-load-trigger-map-context.md
- 07-create-vtc.md → 07-reference-trigger-map.md
Rewritten:
- Phase 4 scenario init: "create VTC" → lightweight "pick from Trigger Map"
- Phase 6 content: load context from Trigger Map, not from VTC
- Templates: "Trigger Map" section → "Business Context" (Phase 1 scope)
Zero VTC references remain in src/.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Suggest workflow: Component Extraction Check prompts user to extract
shared elements when starting the 2nd+ page in a scenario
- Dream workflow: Auto-extracts shared components and reports in summary
- Spec audit Level 3: Cross-Page Duplicate Detection compares sections
across all pages, flags exact/near duplicates with severity levels
- Fix Freya role name (Scenario Architect → Facilitator)
- Remove duplicate trigger-map reference in Eira guide
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Remove duplicate `Reads: 2-strategy/trigger-map.md` line in eira-visual-designer.md
- Rename "Scenario Architect" to "Scenario Facilitator" in freya.activation.md
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Systematic cleanup of Value Trigger Chain (VTC) references in 60 files across
all phases. VTC is replaced with context-appropriate terms:
- "Trigger Map" when referencing the Phase 2 artifact
- "Strategic context" when referencing the extracted business/persona/forces data
- "Value Proposition" for customer-facing value descriptions
Remaining VTC references are only in VTC Workshop definition files (vtc-workshop/,
create-vtc steps) which will be addressed when those workflows are redesigned.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Freya menu: simpler descriptions, logical grouping (SC/UX/SP/SA/GA/DS/DD)
- Freya menu: remove AD and ST, add SC (Scenarios) and SP (Specifications)
- Fix VTC reference in freya agent and workflow-conceptualize
- Conceptualize: save findings to page spec, not separate notes file
- Conceptualize: offer Excalidraw sketching, user sketch first
- Conceptualize: require reading existing specs before drawing
- Conceptualize: no annotations on wireframes rule
- Add Excalidraw tool guide and update tools guide
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
After designing a page, the agent asks what's next: outline and design
the next page in the scenario [N], return to dashboard [R], or validate
[V]. Choosing [N] runs Phase 3's page outline dialog (purpose + exit
action) then designs the page, creating a continuous loop.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
[O] Outline now walks through pages as a conversation: for each page,
ask what the user accomplishes and what interaction moves them forward.
Interactions that leave the page create new steps; interactions that
stay on the page become storyboard items within the page spec.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
[S] → [O] Outline scenario pages. User can cycle through pages
one at a time and stop whenever they want, rather than creating
all at once.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Instead of always creating all page folders, the post-scenario menu
now offers: [N] next scenario, [S] scaffold all page folders,
[D] create first page and design, [C] generate overview. Page folder
structure uses NN.step-slug naming with Sketches/ subfolder.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 4 now checks for existing Phase 3 scenarios before page design.
If no scenario exists, redirects user to Phase 3's 8-question dialog
instead of attempting a lighter inline version.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
After completing a scenario, the user can now choose to:
- [N] Define the next scenario
- [D] Jump to Phase 4 and start designing the first page
- [C] Generate the overview (when all scenarios done)
This lets users design one scenario before defining the rest,
creating a more natural workflow.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Restructure Phase 3 scenario creation around a natural conversation
flow instead of a template checklist:
- 8 strategic questions asked in logical order: transaction → business
goal → user & situation → driving forces → device → starting point
→ best outcome → shortest path
- Two modes: conversation (agent asks, user answers) and suggest
(agent answers from Trigger Map, user reviews)
- Broader transaction definition: includes content browsing and
research, not just purchases/bookings
- Agent creates first page specification after Q8, giving Phase 4
a concrete starting point
- Template reordered to match Q1-Q8 dialog flow
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implement BMad-style IDE handler architecture for WDS:
- Base class (_base-ide.js) with template method pattern
- Dynamic handler discovery via IdeManager
- Priority handlers: Windsurf, Cursor, Claude Code, Cline, GitHub Copilot
- Installer integration after agent compilation step
- Saga activation now branches on starting_point config (pitch vs brief)
- Remove parenthetical hint from learning material prompt
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Remove mimir-orchestrator.agent.yaml (orchestration layer removed)
- Add activation prompt to Saga for immediate greeting
- Saga now direct entry point (no orchestrator needed)
- Greets user by name and starts Product Brief workflow
Workshop finding: 50% bypassed orchestration, Mimir added complexity.
Direct activation simplifies user journey.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- New option: Full Trigger Map vs Simplified VTC vs Skip
- Synced from bmad-method-wds-expansion
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- New entry point that asks Greenfield vs Brownfield upfront
- Prevents users from going through Phase 1-3 with existing codebases
- Adds project configuration questions
- Updates routing to direct brownfield projects to Phase 8
Synced from bmad-method-wds-expansion.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Step 4 was missing a structured menu at the spec review checkpoint.
This caused agents to skip past the approval step without waiting for
explicit user confirmation.
Added:
- Review menu with [y] Approve, [c] Changes, [q] Questions, [a] Advanced Elicitation, [p] Party Mode
- Explicit HALT instruction
- Menu handling section
This aligns step 4 with the menu-driven pattern used in steps 1-3.
Fixes#1304
Also fixes pre-existing prettier issue in src/modules/cis/module.yaml.
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Mårten Angner
Alex Verkhovsky
Brian