feat: BMAD Agent Teams — Claude Code parallel agent orchestration#1
Open
Hidden-History wants to merge 9 commits intomainfrom
Open
feat: BMAD Agent Teams — Claude Code parallel agent orchestration#1Hidden-History wants to merge 9 commits intomainfrom
Hidden-History wants to merge 9 commits intomainfrom
Conversation
Bridge BMAD's 30 agents with Claude Code's native Agent Teams feature (TeamCreate, tmux split-pane teammates, shared task lists, inter-agent messaging). Enables parallel team execution for sprint development, story preparation, test automation, and architecture review. Key components: - .bmad/agent-teams.yaml: Team stage configurations (4 pre-built stages) - .claude/skills/bmad-agent-teams/SKILL.md: Orchestration skill with spawn prompt templates, cost estimation, and lifecycle management - .claude/hooks/scripts/bmad_teammate_idle.py: TeammateIdle quality gate with 2-minute grace period for race condition handling - .claude/hooks/scripts/bmad_task_completed.py: TaskCompleted advisory hook reminding teammates to report status to lead - .bmad/README-agent-teams.md: Setup guide, patterns, and troubleshooting Design decisions: - Star topology (4.4x error amplification vs 17x unstructured) - Single-writer authority (lead exclusively owns sprint-status.yaml) - BMAD slash commands (teammates load personas via workflow engine) - HITL checkpoints (cost estimate + approval before any spawning) - Advisory cost/idle thresholds (Claude Code doesn't expose runtime metrics) Aligns with community direction: bmad-code-org#1550, bmad-code-org#1584, bmad-code-org#1613, bmad-code-org#1628 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add docs/how-to/use-agent-teams.md with step-by-step setup for spawning parallel BMAD agent teams in Claude Code via tmux. Covers prerequisites, stage selection, hook registration, architecture overview, configuration reference, cost management, and troubleshooting. Update README.md to list Agent Teams as a feature alongside Party Mode and add a link in the Documentation section. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The .claude/commands directory was fully gitignored, preventing bmad-team-sprint.md from being committed. Without this slash command, users cannot invoke /bmad-team-sprint to use the Agent Teams feature. Follows the existing .augment/ exception pattern. Related to bmad-code-org#1550. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Remove cost estimator from orchestration flow (Claude Code has no cost API) - Add prerequisite validation system with required/recommended files to all 5 stages - Add research stage for Phase 1 parallel market/domain/technical research - Change story-prep lead from SM to PM for PRD quality alignment - Make lead spawn template generic (stage-agnostic with variable placeholders) - Add Researcher spawn template for 3 research stream types - Document solo-only workflows vs team workflows (with full mapping table) - Fix 14 review issues: agent manifest keys, slash commands, advisory language, permission inheritance, delegate mode, in-process mode, pricing dates, display modes, gitignore, README links Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Sanitize team/teammate names in hook state file paths (path traversal) - Move hook state files from /tmp/ to ~/.claude/tmp/ (world-writable risk) - Fix skill name mismatch: bmad-os-agent-teams → bmad-agent-teams - Add Template Variable Resolution guide to SKILL.md - Add max_teammates enforcement to spawn flow - Broaden lead detection in TaskCompleted hook (*-lead suffix) - Add research stage to slash command description - Document context paths and communication rules in README - Update slash command flow steps (prerequisites, not cost estimate) - Label Architect as lead in architecture-review pattern Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…and review fixes Add /bmad-team-verify command with 5-check dependency analysis (story deps, file ownership, cross-cutting concerns, producer-consumer contracts, quality gate). Split SKILL.md into 3 conditionally-loaded files (SKILL.md hub + verify-workflow.md + sprint-workflow.md) reducing per-invocation token load by ~93%. Replace 7 redundant worker templates with factory pattern. Add staleness check to idle hook. Document experimental API assumptions in hooks. Changes across 10 files: - .claude/skills/bmad-agent-teams/SKILL.md: rewritten as mode-selection hub (813→63 lines) - .claude/skills/bmad-agent-teams/verify-workflow.md: new, 5-check verify flow - .claude/skills/bmad-agent-teams/sprint-workflow.md: new, 13-step orchestration with factory templates - .claude/commands/bmad-team-verify.md: new slash command entry point - .bmad/agent-teams.yaml: parallel-groups in recommended_files, lead responsibilities - .bmad/README-agent-teams.md: dependency verification docs, updated file inventory - .claude/hooks/scripts/bmad_teammate_idle.py: schema docs, path assumption docs, staleness check - .claude/hooks/scripts/bmad_task_completed.py: schema documentation - .gitignore: added verify command exception - docs/how-to/use-agent-teams.md: verify section, story-prep prereq note Reviewed by: 2x Opus + 2x Sonnet parallel review teams (0 issues remaining) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…Ollama/GLM) Fix Step 7 model parameter not being passed to Task tool — all teammates were spawning with default model regardless of config. Add haiku as third valid model tier. Document provider configuration for Ollama Local, Ollama Cloud, and GLM/Z.AI with env var setup and capability matrix. 12 edits across 5 files. 2 review rounds (15 issues found and fixed). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Ollama Cloud URL was `https://ollama.com/api` in 3 files but the verified working URL (SDK spike test, 2026-02-17) is `https://ollama.com`. Also fixes Step 2.5 provider detection pattern to match the correct URL. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
My mistake - I followed this link from an issue thread and thought this was a PR in the BMad Repo :) We should discuss this feature in discord as this is a massive add planned for post v6 many people are considering. But this is a good start - did you start a thread already? |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Add BMAD Agent Teams support for Claude Code's native Agent Teams feature. Enables spawning parallel teams of BMAD agents in tmux panes for sprint development, story preparation, test automation, architecture review, and research.
12 files changed, +1976 lines, 9 commits on
feature/agent-teams-configWhat's Included
New Files
.bmad/agent-teams.yaml— Team compositions for 5 stages (sprint-dev, story-prep, test-automation, architecture-review, research).bmad/README-agent-teams.md— Full configuration reference, troubleshooting, and multi-provider setup.claude/skills/bmad-agent-teams/SKILL.md— Skill hub with conditional loading.claude/skills/bmad-agent-teams/sprint-workflow.md— 13-step spawn & lifecycle workflow.claude/skills/bmad-agent-teams/verify-workflow.md— 5-check dependency verification.claude/commands/bmad-team-sprint.md—/bmad-team-sprintslash command.claude/commands/bmad-team-verify.md—/bmad-team-verifyslash command.claude/hooks/scripts/bmad_teammate_idle.py— Quality gate hook (idle detection).claude/hooks/scripts/bmad_task_completed.py— Quality gate hook (completion reporting)docs/how-to/use-agent-teams.md— User-facing how-to guideREADME.md— Agent Teams bullet in features list.gitignore— Exception for slash command markdown filesArchitecture
5 Pre-Built Stages
sprint-devstory-preptest-automationarchitecture-reviewresearchMulti-Provider Support
https://ollama.comwith GLM-5:cloudQuality
Related Community Issues
Test Plan
Co-Authored-By: Claude Opus 4.6 noreply@anthropic.com