blueprint-dev

Remove losing A/B variant, promote winner, clean up feature flags and tracking code

Analyze A/B test results, determine statistical significance, recommend winner, produce cleanup plan

Check status of all active A/B tests — shows test names, variants, duration, and flag status

Architecture robustness check with parallel security, performance, and data integrity reviews. Produces ADR.

Parallel codebase-wide changes using worktrees — wraps built-in /batch with project conventions, stack profile, and batch manifests

Implement feature with trunk-based development practices — short-lived branch, feature flags, small PRs

Document a solved problem for compound knowledge — 5 parallel agents extract context, solution, cross-references, prevention, and classification

Create 2-3 production A/B design variants with feature flags, tracking, and test plan

Detect project stack, profile technologies, and suggest CLAUDE.md improvements

Record a video walkthrough of a feature and add it to the PR description

Lightweight plan+build for small-to-medium work — auto-triages scope, plans inline, implements with relaxed TBD, suggests /simplify

End-to-end development pipeline — discover, plan, design, architect, build, review, ship, compound — with approval gates

Deep planning with requirements decomposition, research, and scope guard

Multi-agent parallel code review — quality, tests, TBD compliance, and pattern analysis

Final gates check and merge to trunk — TBD compliance, CI status, PR creation

Architecture team swarm — parallel security, performance, data integrity, and core architecture review

Compound knowledge team swarm — parallel problem analysis, solution capture, cross-referencing, prevention, and classification

Design team swarm — variant generation, critique, and A/B test setup

Full swarm — sequential design, architecture, and review teams with approval gates

Review team swarm — parallel code quality, test coverage, TBD compliance, and pattern analysis

Run browser tests on pages affected by current PR or branch

Sets up feature flags, analytics tracking, and the A/B test infrastructure for design variants. Auto-detects flag provider from stack profile.

Designs core architecture and authors Architecture Decision Records (ADRs) in MADR format. Evaluates trade-offs, proposes patterns, and documents rationale.

Classifies solved problems with validated YAML frontmatter — category, severity, module, tags — for searchability and pattern detection.

Compares stack profile against existing CLAUDE.md and suggests targeted additions. Never auto-writes — stages suggestions for user review.

Reviews code for style consistency, clean code principles, design patterns, and project convention adherence. Stack-adaptive — adjusts standards based on detected framework.

Extracts problem metadata from the conversation — module, symptom, environment, timeline, error messages, and affected areas.

Reviews database design, migrations, constraints, transaction boundaries, and data model integrity. Ensures data cannot reach an invalid state.

Evaluates A/B design variant approaches on UX design principles, accessibility, usability heuristics, and hypothesis quality. Does not pick a winner — that's data's job.

Interprets A/B test results, determines statistical significance, recommends a winner, and produces a cleanup plan for the losing variant.

Creates 2-3 production-quality A/B design variants as real code in the source tree. Each variant differs in meaningful UX dimensions and follows project conventions.

Manages feature flag lifecycle — creation, naming, rollout configuration, cleanup, and documentation. Auto-detects the flag provider from stack profile.

Pragmatic implementor for small-to-medium changes — relaxed TBD, proportional testing, optional branching. Used by /bp:go for fast-lane development.

Identifies anti-patterns, SOLID violations, code smells, and refactoring opportunities in code changes. Distinguishes between acceptable pragmatism and genuine problems.

Analyzes architecture and code for N+1 queries, missing indexes, caching opportunities, bundle size impact, bottlenecks, and scalability concerns.

Documents how to prevent recurrence of a solved problem — tests, linting rules, type guards, monitoring alerts, process changes.

Searches docs/solutions/ for similar past problems and finds cross-references in the compound knowledge base.

Decomposes feature requests into structured, testable requirements with acceptance criteria. Identifies ambiguities, dependencies, and risks.

Investigates prior art, existing solutions in the codebase, compound knowledge docs, libraries, and community approaches before building anything new.

Reviews plans and requirements for scope creep, over-engineering, and unnecessary complexity. Guards the boundary between what was requested and what is being planned.

Reviews architecture and code for OWASP Top 10 vulnerabilities, authentication flaws, injection risks, XSS vectors, and authorization gaps.

Captures the root cause and solution from the conversation — what was wrong, why, what fixed it, and what code changed.

Detects project stack, frameworks, tooling, conventions, and project structure. Produces a comprehensive stack profile for other agents to adapt their behavior.

Orchestrates multi-agent team workflows — manages agent spawning, sequencing, result aggregation, and conflict resolution across predefined team swarms.

Analyzes test completeness against requirements, identifies coverage gaps, missing edge cases, and test quality issues.

Enforces trunk-based development compliance — checks branch age, PR size, CI status, feature flag usage, and merge readiness.

Implements features following trunk-based development practices — short-lived branches, small PRs, feature flags, incremental delivery.

Production A/B testing lifecycle for design variants. Covers hypothesis formation, feature flag integration, analytics tracking, statistical analysis, and cleanup.

Browser automation using Vercel's agent-browser CLI. Use when you need to interact with web pages, fill forms, take screenshots, or scrape data. Uses Bash commands with ref-based element selection. Triggers on "browse website", "fill form", "click button", "take screenshot", "scrape page", "web automation".

Multi-agent architecture review combining core architecture design with parallel security, performance, and data integrity assessments. Produces ADRs in MADR format.

Reference for how the built-in /batch command integrates with blueprint-dev workflows — parallel codebase-wide changes using worktrees with project context.

Analyzes detected stack profiles and suggests targeted CLAUDE.md improvements. Never auto-writes to CLAUDE.md — stages suggestions for user review.

Problem documentation methodology for compounding team knowledge. Captures solved problems with structured metadata for searchability, pattern detection, and prevention.

Git worktree management for isolated parallel development. Handles creating, listing, switching, and cleaning up worktrees with interactive confirmations, automatic .env copying, and .gitignore management. Follows KISS principles.

Lightweight planning methodology for small-to-medium work — scope triage, adaptive process, and right-sized ceremony for bug fixes, UI tweaks, and medium features.

Structured planning methodology for decomposing features into testable requirements, researching approaches, and guarding scope. Templates for PRD, spec, and scope documents.

Reference for how the built-in /simplify command integrates with blueprint-dev workflows — when it runs, what it checks, and how it complements review agents.

Detects project technology stacks by analyzing package manifests, config files, directory structures, and code patterns. Produces a structured stack profile for agent adaptation.

Orchestration patterns for multi-agent team workflows — team definitions, sequencing rules, result aggregation, and conflict resolution.

Trunk-based development practices with short-lived branches, feature flags, small PRs, and continuous integration. Enforced via hooks and agents.

1 hook across 1 event