Julius Brussee e637c43402 T-001: Initialize Go module with TUI dependencies
Set up github.com/julb/blueprint-monitor Go module with bubbletea,
lipgloss, bubbles, and creack/pty. Minimal main.go entry point in
cmd/blueprint/.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-17 12:53:36 +01:00

Blueprint

Claude Code plugin + parallel agent launcher for blueprint-driven development with automated iteration loops.

Install

git clone https://github.com/JuliusBrussee/sdd-os.git ~/.blueprint
cd ~/.blueprint
./install.sh

This will:

  1. Register the Blueprint plugin with Claude Code
  2. Install the blueprint CLI command globally
  3. Make all scripts executable

Terminal: parallel agent launcher

blueprint --monitor                     # interactive picker → launch agents in tmux
blueprint --monitor --expanded          # one tmux window per agent with dashboards
blueprint --status                      # check progress from any terminal
blueprint --analytics                   # trends across cycles
blueprint --kill                        # stop everything, clean up worktrees

Default mode (--monitor)

Interactive multi-select picker shows all build sites:

  • Available — ready to launch (pre-selected)
  • In Progress — select to resume from existing worktree
  • Done — struck through (archived sites)

Selected sites each get:

  • Their own git worktree (branch: blueprint/<site-name>)
  • A tmux pane running Claude Code with /bp:build
  • Auto-layout: horizontal for 2-3 agents, tiled for 4+
  • Live status bar showing per-site progress

Staggered launch (5s between agents) to avoid API rate limits.

Expanded mode (--monitor --expanded)

One tmux window per site with the full 3-pane layout:

  • Left (70%) — Claude Code running /bp:build
  • Top-right — live progress: tasks done, tiers, progress bar
  • Bottom-right — live activity: iteration log, git commits

Switch between windows with Ctrl-b <number>.

Analytics (--analytics)

Parses loop logs across all cycles and worktrees:

  • Iterations to convergence per cycle
  • Task outcomes (done/partial/blocked)
  • Failure patterns and dead ends
  • Tier distribution
  • Completion velocity (tasks/iteration, success rate)

Claude: slash commands

/bp:draft       →  draft blueprints (the WHAT)
/bp:architect   →  generate build site (the ORDER)
/bp:build       →  ralph loop (the BUILD)
/bp:inspect     →  gap analysis + peer review (the CHECK)
/bp:merge       →  blueprint-aware branch integration (the SHIP)

1. Draft — write blueprints

/bp:draft                       # interactive — asks what to build
/bp:draft context/refs/         # from PRDs, API docs, research
/bp:draft --from-code           # from existing codebase

Decomposes your project into domains. Each domain gets a blueprint with R-numbered requirements and testable acceptance criteria.

2. Architect — generate build site

/bp:architect                   # all blueprints
/bp:architect --filter v2       # only v2 blueprints

Reads blueprints, breaks requirements into tasks, maps dependencies, organizes into tiers.

3. Build — run the loop

/bp:build                       # implement everything
/bp:build --peer-review         # add Codex (GPT-5.4) review
/bp:build --max-iterations 30

Each iteration: read site → find next unblocked task → read blueprint → implement → validate → commit → loop.

4. Inspect — post-loop check

/bp:inspect                     # gap analysis + peer review

5. Merge — blueprint-aware branch integration

/bp:merge                       # merge all blueprint/* branches into main

After parallel execution, each site lives on its own blueprint/<name> branch. /bp:merge integrates them back into main:

  1. Surveys all branches — commits, file overlaps, dependency order
  2. Reads the blueprints and impl tracking for each branch
  3. Merges in order: infrastructure → features → UI
  4. Resolves conflicts by understanding what each blueprint intended — keeps all features from all branches
  5. Validates after each merge (build, tests, blueprint requirements)
  6. Cleans up worktrees and branches

File structure

context/
├── blueprints/         # Blueprints (persist across cycles)
│   ├── blueprint-overview.md
│   └── blueprint-{domain}.md
├── sites/              # Build sites (one per plan)
│   ├── build-site-ui-v2.md
│   └── archive/        # Completed sites
├── impl/               # Progress (archived between cycles)
│   ├── impl-{domain}.md
│   ├── loop-log.md
│   └── archive/
└── refs/               # Reference materials

All commands

Command Description
/bp:draft Draft blueprints
/bp:architect Generate build site
/bp:build Ralph Loop implementation
/bp:inspect Gap analysis + peer review
/bp:merge Blueprint-aware branch integration
/bp:progress Check site progress
/bp:gap-analysis Compare built vs intended
/bp:revise Trace manual fixes to blueprints
/bp:help Show usage
CLI Description
blueprint --monitor Interactive picker → parallel agents in tmux
blueprint --monitor --expanded One window per agent with dashboards
blueprint --status Check site progress
blueprint --analytics Trends across cycles
blueprint --merge Shows branches ready to merge (use /bp:merge in Claude)
blueprint --kill Stop all agents, clean worktrees

Example

See example.md for full sample conversations.

License

MIT

S
Description
caveman: Caveman encoding for SPEC.md and spec-adjacent writes. Loaded by /spec, /build, /check. Cuts tokens ~75% vs prose while staying precise. Triggers on any…; build: Plan-then-execute implementation against SPEC.md. Native single-thread loop, no sub-agents. On test or build failure, auto-invokes the backprop skill before…; spec: Create, amend, or backprop bugs into SPEC.md at repo root. Sole mutator of the project spec. Triggers when the user asks to write a spec, start a new spec,…; check…
Readme MIT 4.1 MiB
Languages
Markdown 100%