bmad-code-org/bmad-method
40 skills11.7k installs2M starsGitHub
Install
npx skills add https://github.com/bmad-code-org/bmad-methodSkills in this repo
1Bmad Agent Pmbmad-agent-pm is the product-manager persona in the BMAD Method Module Ecosystem, an open-source agile framework installed via `npx bmad-method install` for Claude Code, Cursor, and similar AI IDEs. The PM agent collaborates on refining scope, drafting epics and user stories, prioritizing and sequencing work, and keeping implementation tied to product goals instead of ad-hoc prompts. BMAD Method ships 34+ workflows and 12+ specialized agents (PM, Architect, Developer, UX, and others) with scale-adaptive planning depth. Reach for bmad-agent-pm when a feature idea needs structured backlog artifacts before coding, when sprint scope is drifting from stated goals, or when you want facilitated PM guidance rather than one-shot story generation.397installs2Bmad Agent Architectbmad-agent-architect is an AI & Agent Building skill from bmad-code-org/bmad-method that guides developers through designing agent architectures using BMAD patterns. The workflow covers selecting tools, structuring memory, defining orchestration between steps, and drawing clear agent boundaries so multi-step systems stay maintainable and testable. Developers reach for bmad-agent-architect when moving from a single-shot prompt chain to a production agent with multiple capabilities and failure modes. The skill emphasizes patterns that keep orchestration explicit and boundaries enforceable, which reduces debugging cost as agent complexity grows. Use it during greenfield agent design or when refactoring a brittle prototype into a layered architecture.362installs3Bmad Code Reviewbmad-code-review is a bmad-code-org/BMAD-METHOD implementation skill that orchestrates elite code review inside coding agents. Activation loads `_bmad/bmm/config.yaml`, merges `customize.toml` overrides, and executes a four-step step-file workflow: gather context, launch parallel reviewers, triage findings, and present results. Step 2 launches Blind Hunter via `bmad-review-adversarial-general`, Edge Case Hunter via `bmad-review-edge-case-hunter`, and optionally Acceptance Auditor when a spec file exists, then collects Markdown findings for precision triage. Developers reach for bmad-code-review when they say "run code review" or "review this code" and need adversarial coverage across requirements, architecture, security, style, and test gaps instead of a single-pass summary. The workflow enforces sequential step files, checkpoint halts, and customization through team and user TOML layers under `_bmad/custom/`.360installs4Bmad Create Prdbmad-create-prd is a BMAD Method agent skill that turns rough product ideas into a structured Product Requirements Document covering goals, target users, functional requirements, success metrics, constraints, and release scope. The skill walks an AI coding agent through BMAD-style PRD sections so outputs are consistent enough for design, engineering, and stakeholder review without ad-hoc markdown templates. Developers reach for bmad-create-prd when kicking off a feature, MVP, or internal tool and need a single source of truth that downstream skills or teammates can implement against. It fits early planning in SaaS, API, and agent-backed products where ambiguous scope causes rework. Pair it with architecture or implementation skills after the PRD is approved rather than using it mid-refactor.353installs5Bmad Create Architecturebmad-create-architecture is a BMAD Method solutioning skill in bmad-code-org/BMAD-METHOD that walks developers and agents through collaborative architecture discovery so implementation stays consistent. The legacy workflow used micro-file steps with append-only document building, user approval gates, and initialization via step-01-init.md after resolving customization with resolve_customization.py. BMAD-METHOD v6.9 replaced the original eight-step flow, described in project history as roughly 2,400 lines producing a monolithic architecture.md, with bmad-architecture and an ARCHITECTURE-SPINE.md source of truth. The bmad-create-architecture skill now acts as a deprecated compatibility shim that forwards create intent to bmad-architecture while preserving trigger phrases like create technical architecture. Developers still invoking bmad-create-architecture get guided solution design covering components, data flows, integrations, and key technical decisions, but new work should prefer bmad-architecture directly.351installs6Bmad Helpbmad-help is a bmad-code-org/bmad-method orientation skill that answers BMAD questions and recommends next steps. It analyzes the user's query and current state so developers know which module and phase they are in, what is already complete, and which skill to invoke next. It can augment answers with remote module documentation when broader questions arise. Use it at session start, after context switches, or whenever you are unsure which BMAD command fits the remaining work on a product or agent build.341installs7Bmad Create Epics And Storiesbmad-create-epics-and-stories is a BMAD Method workflow skill in bmad-code-org/bmad-method that transforms PRD requirements and architecture decisions into comprehensive epics and user stories organized by user value. The workflow loads project config from _bmad/bmm/config.yaml, validates prerequisites, then follows stepped instructions to produce epic files embedding stories, acceptance criteria, dependencies, and technically informed implementation notes informed by database, API, and stack choices from architecture. BMAD positions epic creation after architecture so stories reflect real technical constraints rather than PRD-only guesses. Developers and PMs invoke it via /bmad-bmm-create-epics-and-stories with the PM agent after PRD and architecture artifacts exist. Reach for bmad-create-epics-and-stories when a validated requirements set needs backlog decomposition before sprint planning, estimation, or bmad-dev-story implementation workflows.328installs8Bmad Review Adversarial Generalbmad-review-adversarial-general is a bmad-code-org/BMAD-METHOD core skill that performs cynical adversarial review on any submitted artifact. The agent adopts a skeptical reviewer persona, loads content from a diff, branch, uncommitted changes, spec, story, or document, and executes a three-step flow: receive content, adversarial analysis, and present findings. The skill requires finding at least ten issues to fix or improve, looking for what is missing—not only what is wrong—and outputs a Markdown list of descriptions without severity rankings. bmad-code-review invokes this skill as the Blind Hunter layer during parallel code review. Developers reach for bmad-review-adversarial-general when they request a critical or cynical review of plans, specs, or implementations and need explicit weak-assumption surfacing before merge or release.324installs9Bmad Technical Researchbmad-technical-research is a native BMad Method analysis skill that guides agents through structured technical investigation before architecture or PRD work. The workflow spans seven steps: quick topic discovery, initialization, technical overview, integration patterns, architectural patterns, implementation research, and synthesis into a markdown report saved under research/technical with comparison matrices, stack recommendations, performance notes, and citations. Developers invoke it by name (bmad-technical-research) during Phase 1 analysis when evaluating frameworks, comparing tools, or documenting tradeoffs for a feature or platform choice. It requires web search capabilities and respects project TOML customizations for language and output directories. Use it when informal chat comparisons are insufficient and you need a repeatable, sourced technical brief for downstream PRD or architecture skills.324installs10Bmad Agent Analystbmad-agent-analyst is the BMAD Method analyst persona skill that loads Mary, a strategic business analyst focused on evidence-based requirements and discovery workflows. Once activated, the agent greets by name, prefixes messages with its icon, and presents a numbered menu of seven triggers: BP brainstorm project, MR market research, DR domain research, TR technical research, CB create brief, WB working-backwards PRFAQ challenge, and DP document project. The skill resolves customization from customize.toml layers, loads project config from _bmad/bmm/config.yaml, and dispatches matched menu codes to underlying bmad-* workflow skills. Developers reach for bmad-agent-analyst at project kickoff, brownfield documentation, or when stakeholders have vague needs that must become specs before bmad-agent-pm writes a PRD. Invoke by skill name bmad-agent-analyst or ask to talk to Mary. The analyst pairs with foundation discovery workflows and hands off structured briefs to PM and architect agents in the BMAD agile suite.320installs11Bmad Agent Ux Designerbmad-agent-ux-designer is a skill in the bmad-code-org/bmad-method framework for designing conversational and multimodal user experiences around AI agents. It guides developers through agent interaction flows, conversation states, affordances that signal capability boundaries, and error recovery patterns so agent UIs feel trustworthy and usable rather than opaque or brittle. The skill addresses how users discover what an agent can do, how multi-turn context is surfaced, and how failures are communicated without breaking user confidence. It fits the BMAD Method agent-building workflow where specialized skills handle distinct phases of product delivery. Developers reach for bmad-agent-ux-designer when planning chat interfaces, voice or multimodal agent surfaces, or copilot-style embedded assistants and need structured UX decisions before writing frontend or agent orchestration code.318installs12Bmad Brainstormingbmad-brainstorming is a core BMAD Method workflow skill for generating diverse ideas through interactive creative techniques before validation or build. It sets up a session from a topic or problem statement, loads ideation methods from a technique library—including frameworks like SCAMPER and reverse brainstorming—and guides technique-by-technique generation toward 100+ ideas. An anti-bias protocol shifts creative domain every ten ideas to prevent clustering. Session outputs include a self-contained brainstorm.html keepsake, an optional brainstorm-intent.md for downstream BMAD skills, and a .memlog.md session record. Developers reach for bmad-brainstorming when starting a new project, feeling stuck on ideas, or needing structured creativity before product briefs or PRDs. Choose interaction stances: Facilitator, Creative Partner, or Ideate-for-me. Pair results with bmad-spec or bmad-product-brief for the next BMAD lifecycle steps.318installs13Bmad Document Projectbmad-document-project is a bmad-code-org/bmad-method skill triggered as DP on Analyst or Technical Writer agents to document brownfield projects for AI-assisted development. The workflow classifies repos against 12 project types, detects architecture from package.json and folder layout, scans entry points, and generates index.md, project-overview.md, architecture.md, source-tree-analysis.md, component-inventory.md, and optional deep-dive files after exhaustive module reads. Modes include initial scan, full rescan after major changes, and deep-dive on APIs, features, or services. Outputs land in the project docs folder as AI-friendly retrieval sources—no application code is written. Developers reach for bmad-document-project when onboarding to legacy repos, refreshing docs post-refactor, or feeding accurate context into BMAD PRDs. With 249 catalog installs, it sits in the BMAD Method suite of 34+ workflows.318installs14Bmad Generate Project Contextbmad-generate-project-context is a native BMAD Method skill that produces project-context.md—an LLM-optimized implementation guide loaded automatically by bmad-create-architecture, bmad-dev-story, bmad-code-review, bmad-quick-dev, and related workflows. The three-step workflow discovers conventions from repository files, generates concise rules agents might miss, and completes a reviewable context document defaulting to _bmad-output/project-context.md. It captures technology stack versions, critical TypeScript or framework rules, testing patterns, and unobvious conventions rather than generic best practices. Brownfield teams run it to prevent agents from breaking established patterns; greenfield teams run it after architecture to encode decisions. Re-run after major refactors or when agent outputs drift from team standards.318installs15Bmad Product Briefbmad-product-brief is a workflow skill from the BMad Method (BMM) analysis module that coaches developers through creating, updating, or validating a product brief before PRD or architecture work. Installed via npx bmad-method install, the skill registers in IDE skill directories such as .claude/skills/ or .agents/skills/ and runs a conversational Discovery phase with Fast path ([ASSUMPTION] tags) or Coaching path (pushback on thin answers). It writes brief.md with YAML frontmatter, optional addendum.md for depth that belongs downstream, and .memlog.md as an append-only decision log via memlog.py. Headless mode completes create, update, or validate intents with JSON status output listing artifact paths. External handoffs can route finished briefs to Confluence or Notion through configured MCP tools. Reach for bmad-product-brief when a product concept is clear enough to document but not yet ready for a full PRD.318installs16Bmad Agent Devbmad-agent-dev loads Amelia, the BMAD Method senior software engineer persona focused on test-first story execution and verified delivery. Activation resolves customize.toml layers via resolve_customization.py, loads persistent facts and _bmad/bmm/config.yaml, then presents a capabilities table with seven triggers: DS dev-story writes tests and code for the next story, QD quick-dev handles ad-hoc implementation, QA generates API and E2E tests, CR runs multi-facet code review, SP sprint planning, CS create story, and ER epic retrospective. Amelia enforces ordered task execution, marks subtasks complete only when tests pass, and documents changes in story Dev Agent Record sections. Developers reach for bmad-agent-dev during sprint implementation after epics and stories exist, or when they ask to talk to Amelia. The skill stays in character with icon-prefixed messages until dismissed and dispatches exact registered bmad-* workflow skills on menu match without inventing capabilities.317installs17Bmad Create Storybmad-create-story is a BMAD Method implementation workflow skill that turns epics, PRD, architecture, and UX artifacts into a dedicated story file an agent can implement without guessing. The skill loads sprint-status.yaml, discovers the next backlog story, exhaustively analyzes planning documents, and writes story-[slug].md with acceptance criteria, tasks, dev notes, and sizing hints while updating status to ready-for-dev. Developers reach for bmad-create-story when a sprint plan exists but the next backlog item still lacks the implementation context an LLM developer needs. It pairs with bmad-dev-story and code-review inside the BMAD 34+ workflow ecosystem. Trigger phrases include create the next story or create story [identifier].315installs18Bmad Dev Storybmad-dev-story is a BMAD Method implementation workflow that reads a context-filled story-[slug].md, discovers the first ready-for-dev item in sprint-status.yaml, and executes every task in order using red-green-refactor until acceptance criteria pass. The skill captures baseline_commit in story frontmatter, marks tasks complete only after tests pass, updates the File List and Dev Agent Record, and moves story status to review. Developers reach for bmad-dev-story after bmad-create-story when a story is ready-for-dev and an agent should implement without pausing at session boundaries. It resumes cleanly after code-review follow-ups and recommends code-review with a different LLM when finished. HALT conditions block progress on missing config or unapproved dependencies.315installs19Bmad Sprint Planningbmad-sprint-planning is a bmad-code-org/bmad-method skill that runs agile sprint planning rituals within the BMAD spec-driven framework. It decomposes epics into sized user stories, assigns priorities, and aligns team capacity before agents or engineers begin implementation work. The skill sits between validated PRDs and execution workflows like bmad-quick-dev, ensuring backlog items are right-sized and sequenced for the upcoming iteration. BMAD Method provides scale-adaptive planning across 34+ workflows with specialized agents for PM, architect, and developer roles. Reach for bmad-sprint-planning at iteration boundaries when epics need story breakdown, priority ordering, or capacity checks against available agent and engineer bandwidth. Outputs become the sprint backlog that quick-dev sessions or manual implementation can consume, keeping BMAD planning artifacts synchronized with what actually ships each cycle.311installs20Bmad Quick Devbmad-quick-dev is a bmad-code-org/bmad-method skill implementing the BMAD Quick Flow track for small features and bug fixes that need a technical specification but not a full PRD and architecture cycle. The unified workflow handles planning and implementation in one focused agent session, producing working code, tests, and docs from scoped stories. BMAD Method scales planning depth to project complexity—the Quick Flow track contrasts with the standard Method track requiring PRDs, UX specs, and architecture before coding. Reach for bmad-quick-dev when a story is already scoped, the change is bounded, and you want BMAD rituals without enterprise planning overhead. The skill suits bug fixes, incremental features, and refactors where a tech spec plus agent execution beats multi-document planning. Outputs feed directly into repository commits ready for review gates like codex-review or standard PR checks.310installs21Bmad Party Modebmad-party-mode is the BMAD Method multi-agent orchestrator invoked with `/bmad-party-mode` to gather installed persona agents—PM, Architect, Dev, UX Designer, QA, and module-specific roles—into one facilitated discussion. The BMad Master selects relevant agents per message, manages turn-taking, and supports four execution modes: session (default single-model roleplay), auto, subagent (independent agents), and agent-team (persistent roundtable). Flags include `--mode`, `--non-interactive`, `--party` for saved casts, and `--solo` to force single-model play. Developers steer with follow-ups, challenges, and topic changes until satisfied. Use bmad-party-mode for complex architecture decisions, epic retrospectives, or PR reviews needing conflicting expert lenses—not for single-agent boilerplate generation or unattended CI tasks.309installs22Bmad Editorial Review Prosebmad-editorial-review-prose is a clinical copy-editing skill from the BMAD Method that reviews PRDs, READMEs, ADRs, and user-facing copy for communication issues that impede comprehension. The skill applies Microsoft Writing Style Guide principles as a baseline, skips code blocks and frontmatter, and never rewrites for stylistic preference—only genuine clarity problems. It accepts required content input plus optional style_guide and reader_type parameters (humans or llm), runs a 4-step STEPS workflow, deduplicates repeated issues, and outputs a three-column markdown table with Original Text, Revised Text, and Changes—or reports no editorial issues identified. Developers reach for bmad-editorial-review-prose before shipping documentation when prose clarity matters more than structural reorganization.308installs23Bmad Market Researchbmad-market-research is a BMAD Method skill for structured pre-build discovery. It collects market size estimates, competitor positioning, pricing signals, and buyer pain points so teams can validate direction before writing production code. Developers and PMs reach for bmad-market-research when an agent-assisted brainstorm needs evidence-backed scope instead of guesswork feature lists. The skill outputs research artifacts that inform MVP cuts, pricing hypotheses, and positioning statements used in subsequent validate and build phases within the BMAD workflow.306installs24Bmad Qa Generate E2e Testsbmad-qa-generate-e2e-tests is a built-in QA Automate workflow in the BMAD Method BMM module that turns shipped features into executable test suites. The skill follows five ordered steps: detect the project's test framework from package.json and existing tests (Jest, Vitest, Playwright, Cypress, or a suggested runner), identify target features from prompts or codebase discovery, generate API tests covering status codes and response shape, generate E2E tests with semantic locators and visible-outcome assertions, then run and repair failing specs. Developers invoke it via the Developer agent QA command or the bmad-qa-generate-e2e-tests skill after stories in an epic are implemented and code-reviewed. It focuses on happy paths and one or two critical error cases rather than exhaustive coverage, and defers enterprise test-architecture strategy to the separate TEA module documented in BMAD testing references.306installs25Bmad Retrospectivebmad-retrospective is a BMAD-METHOD implementation skill that runs structured post-epic reviews to extract lessons and assess success. It loads configuration from _bmad/bmm/config.yaml, detects the completed epic from sprint-status.yaml in implementation artifacts, and verifies all stories reach done status before proceeding. The retrospective uses a two-part format—Epic Review plus Next Epic Preparation—with party-mode agent dialogue formatted as Name (Role): dialogue and a no-blame, systems-focused facilitation style. Input files include epic markdown, optional previous retrospectives, architecture docs, and PRD artifacts from planning_artifacts paths. Activation resolves customize.toml layers and persistent facts before greeting the configured user. Developers reach for bmad-retrospective when an epic finishes and the team needs actionable retrospective output instead of informal chat debriefs.305installs26Bmad Sprint Statusbmad-sprint-status is a Claude Code agent skill from bmad-code-org/bmad-method that helps engineering leads track daily sprint progress, blockers, and burndown signals during an active delivery cycle. The skill supports BMAD-method teams that need a lightweight ritual for rebalance decisions—seeing which stories stalled, which dependencies block throughput, and whether remaining work fits the sprint window. Developers and tech leads reach for bmad-sprint-status when standups, async updates, or mid-sprint reviews need structured status instead of ad-hoc chat summaries. The skill focuses on operational visibility during iteration rather than initial roadmap planning or post-release analytics. Use it once a sprint is underway and delivery predictability depends on timely blocker surfacing and work rebalancing.303installs27Bmad Advanced Elicitationbmad-advanced-elicitation is a BMAD Method skill that runs structured second-pass reasoning on LLM-generated content instead of vague retry prompts. The skill presents five context-matched elicitation methods—such as Pre-mortem Analysis, First Principles Thinking, Red Team vs Blue Team, and Socratic Questioning—from a registry of dozens of techniques. Developers pick a method, the agent re-examines output through that analytical lens, and they accept or discard improvements iteratively. Reach for bmad-advanced-elicitation when specs, architecture docs, or agent designs feel shallow and need stress-tested requirements, constraints, and success criteria before build begins. The skill integrates at BMAD workflow decision points after initial content generation.302installs28Bmad Domain Researchbmad-domain-research is a bmad-code-org/bmad-method skill that runs structured domain investigation before developers lock build scope. It examines unfamiliar problem spaces—industry terminology, regulatory constraints, user workflows, and incumbent solutions—to inform product direction within the BMAD spec-driven agile framework. The BMAD Method ecosystem provides 34+ workflows and 12+ specialized agents across analysis, planning, architecture, and implementation phases; domain research sits in the analysis track validating assumptions that later feed PRDs and architecture docs. Reach for bmad-domain-research when entering a new vertical, evaluating feasibility of a feature concept, or comparing how existing products solve the same workflow. The skill produces research artifacts agents and engineers can reference during PRD drafting and sprint planning, reducing hallucinated requirements in downstream BMAD quick-dev or sprint-planning sessions.301installs29Bmad Correct Coursebmad-correct-course is a BMAD Method implementation skill from bmad-code-org/bmad-method invoked when a team says correct course or proposes a sprint change during active delivery. The SKILL.md delegates to workflow.md, where a Scrum Master persona analyzes the triggering issue, loads project-context.md when present, and verifies access to PRD, epics, architecture, and UX specifications before assessing impact. Developers reach for bmad-correct-course when implementation diverges from approved requirements, epics need rewriting mid-sprint, or UX and technical artifacts conflict after new constraints emerge. The workflow supports incremental collaborative refinement or batch review modes and culminates in a structured Sprint Change Proposal covering epic impact, story impact, artifact conflicts, technical implications, and detailed change recommendations. Output targets updated epics, stories, or PRD sections with actionable handoff rather than ad-hoc chat advice.297installs30Bmad Validate Prdbmad-validate-prd is a bmad-code-org/bmad-method skill that audits product requirement documents before any implementation work begins. It checks PRDs for completeness, technical feasibility, and alignment with BMAD spec-driven methodology so downstream sprint planning and quick-dev sessions start from sound requirements. The standard BMAD Method track requires PRDs and architecture before coding, unlike the Quick Flow track for small bug fixes needing only a tech spec. Reach for bmad-validate-prd when a draft PRD exists but needs gap analysis, contradiction checks, or feasibility stress-testing before engineers and agents commit to build scope. The skill acts as a scope gate between domain research and sprint planning, catching missing acceptance criteria, unstated constraints, or misaligned agent workflows early. Outputs include annotated PRD feedback ready for revision or approval.296installs31Bmad Review Edge Case Hunterbmad-review-edge-case-hunter is a method-driven code review skill in bmad-code-org/BMAD-METHOD that acts as a pure path tracer—never judging code quality, only listing missing guards. It accepts a diff, full file, or function as content, optionally honors also_consider hints, and executes five ordered steps: receive content, exhaustive path analysis, completeness validation, deletion check for meaningful removals, and JSON output. Each finding includes location, trigger_condition, guard_snippet, and potential_consequence fields with no severity labels or rankings. When a diff is provided, analysis stays within changed hunks unless external references are explicit. Developers reach for bmad-review-edge-case-hunter orthogonal to adversarial review when they need exhaustive boundary enumeration on specs, handlers, or PR diffs before merge.295installs32Bmad Edit Prdbmad-edit-prd is a thin compatibility shim in bmad-code-org/bmad-method that forwards update-intent PRD work to the consolidated bmad-prd skill. BMAD consolidated bmad-create-prd, bmad-edit-prd, and bmad-validate-prd into one bmad-prd workflow supporting Create, Update, and Validate intents with facilitated discovery, validation checklists, and HTML report rendering. When invoked, bmad-edit-prd resolves legacy customization from _bmad/custom/bmad-edit-prd.toml—activation steps, persistent facts, and on_complete hooks—then hands the conversation to bmad-prd with intent locked to update. Developers still on bmad-edit-prd should migrate to bmad-prd directly for full template, checklist, and external handoff options. The shim exists so existing project overrides keep working until removal in BMAD v7.291installs33Bmad Editorial Review Structurebmad-editorial-review-structure is a BMAD Method skill that acts as a structural editor proposing cuts, reorganization, and simplification while preserving comprehension. The skill targets high-value density: every section must justify its existence, redundancy is treated as failure, and brevity is framed as clarity for scannable technical documents. Developers run bmad-editorial-review-structure before copy editing when a spec, PRD, architecture doc, or handoff document feels bloated, out of order, or hard to scan. It fits build-phase documentation workflows where structural problems must be fixed before wording polish or engineering kickoff.285installs34Bmad Prfaqbmad-prfaq helps teams author PR/FAQ artifacts using the BMAD method, defining customer problems, envisioned solutions, and frequently asked questions to validate scope and messaging before build.257installs35Bmad Checkpoint PreviewBmad-checkpoint-preview exposes upcoming and completed BMAD method checkpoints so product and engineering stakeholders can inspect workflow state, validate artifacts at phase gates, and approve continuation before agents proceed to the next planned deliverable.236installs36Bmad Customizebmad-customize authors and updates customization override files for installed BMad skills. A developer uses it to translate plain-English intent into a correctly placed TOML override under _bmad/custom/, choosing between agent-level and workflow-level surfaces and team versus user placement. It discovers customizable skills, composes sparse overrides, and verifies they resolve correctly.166installs37Bmad PrdThis skill helps a user create, update, or validate a Product Requirements Document. Acting as a facilitator, it runs a discovery flow that starts with a brain dump and calibrates scope and rigor to the user's needs, optionally spawning research subagents. It maintains an append-only memory log of decisions and can run headless, and it can also validate an existing PRD by critique only.166installs38Bmad Uxbmad-ux is a BMAD Method skill that plans UX patterns and design specifications through elicitation. It produces two peer contracts: DESIGN.md for visual identity following the Google Labs design.md spec, and EXPERIENCE.md for information architecture, states, interactions, and accessibility. Developers use it when they want to create UX design or plan the UX for a product.129installs39Bmad Specbmad-spec is a BMAD Method skill that turns any intent input, such as a brain dump, PRD, RFC, brief, or transcript, into a canonical SPEC.md carrying a five-field kernel plus companion files. It derives the spec from an append-only memory log so multiple downstream skills can update the same contract without merge drift. Developers use it when they say 'create a spec', 'distill this into a spec', 'validate this spec', or 'update the spec.'118installs40Bmad Architecturebmad-architecture produces an architecture spine: a lean set of invariants (design paradigm, boundary and dependency rules, state ownership) that keep independently built units from diverging. A developer uses it to design a technical architecture or solution design, guided by a coaching path that elicits and pressure-tests the load-bearing decisions. Decisions are recorded in an append-only memlog and distilled into the final spine.92installs