From 99685319fd672ab398e4a9854208ecb58ac28f4d Mon Sep 17 00:00:00 2001 From: Kieran Klaassen Date: Sun, 25 Jan 2026 19:38:05 -0500 Subject: [PATCH 1/3] feat(commands): add /slfg swarm-based autonomous engineering workflow Add new /slfg command that uses TeammateTool to spawn specialized agents (Planner, Researcher, Worker, Reviewer, Tester, Resolver, Video) that work in parallel on different phases of the development lifecycle. Key features: - Parallel execution: planning + research run simultaneously - Shared task list with dependencies for coordination - Team lead orchestrates and approves plans before implementation - Review + testing run in parallel after implementation - Full autonomous workflow from feature description to PR with video This provides maximum parallelism compared to the sequential /lfg command. Co-Authored-By: Claude Opus 4.5 --- .../2026-01-25-feat-swarm-lfg-command-plan.md | 453 ++++++++++++++ .../.claude-plugin/plugin.json | 4 +- plugins/compound-engineering/CHANGELOG.md | 20 + plugins/compound-engineering/README.md | 15 +- plugins/compound-engineering/commands/slfg.md | 557 ++++++++++++++++++ 5 files changed, 1043 insertions(+), 6 deletions(-) create mode 100644 docs/plans/2026-01-25-feat-swarm-lfg-command-plan.md create mode 100644 plugins/compound-engineering/commands/slfg.md diff --git a/docs/plans/2026-01-25-feat-swarm-lfg-command-plan.md b/docs/plans/2026-01-25-feat-swarm-lfg-command-plan.md new file mode 100644 index 0000000..a77755c --- /dev/null +++ b/docs/plans/2026-01-25-feat-swarm-lfg-command-plan.md @@ -0,0 +1,453 @@ +--- +title: "feat: Add /slfg Swarm-Based Autonomous Engineering Workflow" +type: feat +date: 2026-01-25 +--- + +# feat: Add /slfg Swarm-Based Autonomous Engineering Workflow + +## Overview + +Create a new `/slfg` (Swarm LFG) command that transforms the existing sequential `/lfg` workflow into a fully autonomous, parallel swarm-based engineering pipeline. The swarm uses Claude's TeammateTool to spawn specialized teammates that work concurrently on different phases of the development lifecycle, with a team lead orchestrating the work. + +**Key difference from `/lfg`:** Instead of running commands sequentially (plan → deepen → work → review → resolve → test → video), the swarm spawns specialized agents that work in parallel where possible, coordinate through task lists, and communicate via the TeammateTool messaging system. + +## Problem Statement / Motivation + +The current `/lfg` command runs phases sequentially: +1. `/workflows:plan` (blocking) +2. `/deepen-plan` (blocking) +3. `/workflows:work` (blocking) +4. `/workflows:review` (blocking) +5. `/resolve_todo_parallel` (parallel within phase, but phase is blocking) +6. `/test-browser` (blocking) +7. `/feature-video` (blocking) + +This means: +- **Total time = sum of all phases** - no overlap +- **No concurrent research** - research agents in deepen-plan could run while planning continues +- **Sequential bottlenecks** - waiting on slow phases blocks everything +- **Single context** - one agent holds all context, risking context overflow + +**The swarm approach:** +- **Parallel execution** - Multiple agents work simultaneously +- **Shared task list** - Coordination through structured tasks with dependencies +- **Specialized teammates** - Each agent focuses on their specialty +- **Autonomous operation** - Team lead orchestrates, teammates execute independently +- **Context distribution** - Each agent manages its own context + +## Proposed Solution + +### Architecture: Swarm-Based Engineering Team + +``` +┌─────────────────────────────────────────────────────────────────┐ +│ TEAM LEAD │ +│ (Orchestrates, assigns tasks, approves plans, handles decisions)│ +└─────────────────────────────────────────────────────────────────┘ + │ + ┌───────────────────┼───────────────────┐ + ▼ ▼ ▼ + ┌──────────┐ ┌──────────┐ ┌──────────┐ + │ Planner │ │Researcher│ │ Worker │ + │ Agent │ │ Agent │ │ Agent │ + └──────────┘ └──────────┘ └──────────┘ + │ │ │ + ▼ ▼ ▼ + ┌──────────┐ ┌──────────┐ ┌──────────┐ + │ Reviewer │ │ Tester │ │ Video │ + │ Agent │ │ Agent │ │ Agent │ + └──────────┘ └──────────┘ └──────────┘ +``` + +### Workflow Phases + +```mermaid +graph TD + A[Start /slfg] --> B[Team Lead: Create Team & Tasks] + B --> C[Spawn Planner Agent] + B --> D[Spawn Researcher Agent] + C --> E{Plan Complete?} + D --> F[Research Results Ready] + F --> E + E -->|Yes| G[Team Lead: Approve Plan] + G --> H[Spawn Worker Agent] + H --> I{Implementation Complete?} + I -->|Yes| J[Spawn Reviewer Agent] + I -->|Yes| K[Spawn Tester Agent] + J --> L{Review Complete?} + K --> M{Tests Pass?} + L --> N[Spawn Resolver Agents in Parallel] + M --> N + N --> O{All Resolved?} + O -->|Yes| P[Spawn Video Agent] + P --> Q[Create PR with Video] + Q --> R[Team Lead: Cleanup & Report] +``` + +### Agent Roles + +| Agent | Role | Tools/Skills | Works With | +|-------|------|--------------|------------| +| **Team Lead** | Orchestrate, approve plans, handle decisions | TeammateTool, TaskList, AskUserQuestion | All agents | +| **Planner** | Create implementation plan | workflows:plan, AskUserQuestion | Researcher | +| **Researcher** | Deepen plan with research | deepen-plan, Context7, WebSearch | Planner | +| **Worker** | Implement the plan | Edit, Write, Bash, git | Team Lead | +| **Reviewer** | Review implementation | All review agents (kieran, dhh, security, etc.) | Worker | +| **Tester** | Run browser tests | agent-browser, test-browser skill | Worker | +| **Resolver** | Fix review findings | pr-comment-resolver pattern | Reviewer | +| **Video** | Record feature demo | feature-video skill | Worker | + +### Coordination Mechanisms + +#### 1. Task List (Shared Workspace) +``` +~/.claude/tasks/{team-name}/ +├── 001-pending-p1-create-plan.md +├── 002-pending-p1-research-best-practices.md +├── 003-blocked-p1-implement-feature.md # blockedBy: [001, 002] +├── 004-blocked-p2-review-implementation.md # blockedBy: [003] +├── 005-blocked-p2-run-browser-tests.md # blockedBy: [003] +├── 006-blocked-p3-resolve-findings.md # blockedBy: [004, 005] +└── 007-blocked-p3-record-video.md # blockedBy: [006] +``` + +#### 2. Teammate Messaging +- **Team Lead → Teammate**: Task assignment, plan approval, shutdown requests +- **Teammate → Team Lead**: Idle notification, completion status, blockers +- **Teammate → Teammate**: Direct coordination when needed + +#### 3. Plan Mode for Critical Phases +- **Planner Agent** uses `plan_mode_required: true` +- Team Lead receives plan approval requests +- Plan must be approved before Worker begins + +## Technical Approach + +### Phase 1: Command Setup + +**File:** `plugins/compound-engineering/commands/slfg.md` + +```markdown +--- +name: slfg +description: Swarm-based autonomous engineering workflow +argument-hint: "[feature description]" +--- + +# Swarm LFG - Full Autonomous Engineering + +## Step 1: Create Swarm Team + +Use TeammateTool to create a new team: + +\`\`\` +Teammate operation: spawnTeam +team_name: "slfg-{timestamp}" +description: "Autonomous engineering swarm for: {feature_description}" +\`\`\` + +## Step 2: Create Task List + +Use TaskCreate to create the full task dependency graph: + +1. **Plan Feature** (P1, no dependencies) + - Subject: "Create implementation plan" + - Description: "Run /workflows:plan for: {feature_description}" + +2. **Research & Deepen** (P1, no dependencies - runs parallel with planning) + - Subject: "Research best practices" + - Description: "Run /deepen-plan research agents" + +3. **Implement Feature** (P1, blockedBy: [1, 2]) + - Subject: "Implement the approved plan" + - Description: "Run /workflows:work on the plan file" + +4. **Review Implementation** (P2, blockedBy: [3]) + - Subject: "Review code changes" + - Description: "Run /workflows:review agents" + +5. **Test in Browser** (P2, blockedBy: [3]) + - Subject: "Run browser tests" + - Description: "Run /test-browser on affected pages" + +6. **Resolve Findings** (P2, blockedBy: [4, 5]) + - Subject: "Fix review findings and test failures" + - Description: "Run /resolve_todo_parallel" + +7. **Record Video** (P3, blockedBy: [6]) + - Subject: "Record feature demo" + - Description: "Run /feature-video for PR" + +8. **Create PR** (P3, blockedBy: [7]) + - Subject: "Create pull request" + - Description: "Push and create PR with video" + +## Step 3: Spawn Teammates + +Spawn teammates for parallel phases using Task tool: + +\`\`\` +# Spawn Planner (with plan_mode_required) +Task general-purpose: + team_name: "slfg-{timestamp}" + name: "planner" + mode: "plan" + prompt: "You are the Planner. Claim task #1 (Create implementation plan). + Run /workflows:plan for: {feature_description} + When complete, use ExitPlanMode to request approval." + +# Spawn Researcher (runs in parallel) +Task general-purpose: + team_name: "slfg-{timestamp}" + name: "researcher" + prompt: "You are the Researcher. Claim task #2 (Research best practices). + Research patterns, best practices, and documentation for: {feature_description} + Write findings to docs/research/{feature}.md + Mark task complete when done." +\`\`\` + +## Step 4: Orchestration Loop + +Team Lead monitors and orchestrates: + +1. **Wait for Plan Approval Request** + - Receive `plan_approval_request` message from Planner + - Review the plan + - Use `approvePlan` or `rejectPlan` with feedback + +2. **Spawn Worker After Approval** + - Once plan approved + research complete + - Spawn Worker agent to implement + +3. **Spawn Review + Test in Parallel** + - Once implementation complete + - Spawn Reviewer and Tester simultaneously + +4. **Spawn Resolvers** + - Once review + tests complete + - Spawn resolver agents for each finding (parallel) + +5. **Spawn Video Agent** + - Once all findings resolved + - Record and upload feature demo + +6. **Create PR & Cleanup** + - Create PR with video + - Use `cleanup` operation to remove team resources + - Report final status to user + +## Step 5: Completion + +Output summary: +- PR URL +- Time taken +- Agents spawned +- Tasks completed +``` + +### Phase 2: Teammate Agent Prompts + +Each spawned teammate needs specific instructions: + +#### Planner Agent +```markdown +You are the **Planner** for the slfg swarm. + +Your job: +1. Read your assigned task using TaskGet +2. Run /workflows:plan for the feature +3. Use ExitPlanMode when plan is ready for approval +4. Wait for approval before marking task complete + +You have plan_mode_required enabled. You MUST get approval before implementation begins. + +When you finish or are blocked, send a message to the team lead. +``` + +#### Worker Agent +```markdown +You are the **Worker** for the slfg swarm. + +Your job: +1. Read the approved plan file +2. Implement following /workflows:work patterns +3. Create incremental commits +4. Mark your task complete when implementation is done + +Push to a feature branch, not main. +``` + +#### Reviewer Agent +```markdown +You are the **Reviewer** for the slfg swarm. + +Your job: +1. Run all review agents in parallel against the changes +2. Create todos for each finding in todos/ directory +3. Synthesize findings +4. Mark your task complete + +Do NOT fix the issues - just document them for the Resolver agents. +``` + +### Phase 3: Error Handling & Recovery + +| Scenario | Handling | +|----------|----------| +| Teammate fails | Team Lead receives error, can respawn or handle manually | +| Plan rejected | Planner revises and resubmits | +| Tests fail | Create P1 todos, spawn resolvers to fix | +| Context overflow | Checkpoint/resume with fresh context | +| User interrupt | Cleanup team, preserve work | + +### Phase 4: Progress Visibility + +The team lead maintains visibility: + +```markdown +## Swarm Status: slfg-20260125-123456 + +**Feature:** Add user authentication + +| Task | Status | Agent | Progress | +|------|--------|-------|----------| +| Create plan | Complete | planner | 100% | +| Research | Complete | researcher | 100% | +| Implement | In Progress | worker | 60% | +| Review | Pending | - | - | +| Test | Pending | - | - | +| Resolve | Pending | - | - | +| Video | Pending | - | - | + +**Active Agents:** 1 (worker) +**Messages:** 12 received +**Elapsed:** 4m 32s +``` + +## Acceptance Criteria + +### Functional Requirements +- [ ] `/slfg` command creates a team and task list +- [ ] Planner and Researcher run in parallel at start +- [ ] Plan requires team lead approval before implementation +- [ ] Worker implements after plan approval +- [ ] Reviewer and Tester run in parallel after implementation +- [ ] Resolver agents fix findings in parallel +- [ ] Video agent records demo +- [ ] PR created with all artifacts +- [ ] Team resources cleaned up on completion + +### Non-Functional Requirements +- [ ] Faster than sequential `/lfg` for multi-phase workflows +- [ ] Graceful handling of agent failures +- [ ] Progress visible to user throughout +- [ ] Context distributed across agents (no single-agent overflow) +- [ ] Works with existing commands/skills (composition) + +### Quality Gates +- [ ] All spawned agents complete their tasks +- [ ] No orphaned team resources after completion +- [ ] Plan approval gate enforced +- [ ] Final PR includes video and proper description + +## Dependencies & Prerequisites + +1. **TeammateTool** - Already available in Claude Code +2. **Task tools** (TaskCreate, TaskUpdate, TaskList, TaskGet) - Already available +3. **Existing commands** - /workflows:plan, /workflows:work, etc. +4. **Agent spawning** - Task tool with team_name and name parameters + +## Risk Analysis & Mitigation + +| Risk | Likelihood | Impact | Mitigation | +|------|------------|--------|------------| +| Agent coordination failures | Medium | High | Explicit task dependencies, team lead oversight | +| Context overflow in team lead | Medium | Medium | Delegate work, keep lead context light | +| Deadlock from circular dependencies | Low | High | Linear task dependency graph | +| Orphaned agents | Medium | Low | Shutdown requests, timeout cleanup | +| Plan approval bottleneck | Medium | Medium | Clear approval criteria, fast feedback | + +## Implementation Phases + +### MVP (Phase 1) +- [ ] Create `/slfg` command with basic team creation +- [ ] Implement task list creation with dependencies +- [ ] Spawn Planner + Researcher in parallel +- [ ] Plan approval flow with team lead +- [ ] Spawn Worker after approval +- [ ] Basic completion and cleanup + +### Enhanced (Phase 2) +- [ ] Parallel Reviewer + Tester spawning +- [ ] Dynamic Resolver agent spawning +- [ ] Video agent integration +- [ ] Progress dashboard output +- [ ] Error recovery and retry + +### Polished (Phase 3) +- [ ] Checkpoint/resume for interrupted swarms +- [ ] Cost tracking per agent +- [ ] Performance metrics (time savings vs /lfg) +- [ ] User configuration options + +## Files to Create/Modify + +| File | Action | Description | +|------|--------|-------------| +| `commands/slfg.md` | Create | Main swarm command | +| `skills/swarm-engineering/SKILL.md` | Create | Skill with teammate prompts | +| `README.md` | Update | Add /slfg to command list | +| `CHANGELOG.md` | Update | Document new feature | +| `.claude-plugin/plugin.json` | Update | Bump version, update counts | + +## References & Research + +### Internal References +- `/lfg` command: `plugins/compound-engineering/commands/lfg.md` +- Agent-native architecture: `plugins/compound-engineering/skills/agent-native-architecture/SKILL.md` +- Execution patterns: `plugins/compound-engineering/skills/agent-native-architecture/references/agent-execution-patterns.md` +- Parallel resolution: `plugins/compound-engineering/commands/resolve_parallel.md` + +### TeammateTool Operations +From system prompt: +- `spawnTeam` - Create a new team +- `write` - Send message to ONE teammate +- `broadcast` - Send to ALL teammates (use sparingly) +- `approvePlan` / `rejectPlan` - Handle plan approval requests +- `requestShutdown` / `approveShutdown` - Graceful shutdown +- `cleanup` - Remove team and task directories + +### Task Tool Integration +- `team_name` parameter to join spawned agents to team +- `name` parameter to give agents identifiable names +- `mode: "plan"` for plan_mode_required behavior + +## Alternative Approaches Considered + +### Alternative 1: Enhanced Sequential (Rejected) +Keep `/lfg` sequential but add more parallelism within phases. +- **Rejected because:** Doesn't leverage full swarm capabilities, still fundamentally sequential. + +### Alternative 2: Background Agents Only (Rejected) +Use `run_in_background` instead of teams for parallelism. +- **Rejected because:** No coordination mechanism, no shared task list, harder to monitor. + +### Alternative 3: Hybrid Sequential/Parallel (Considered) +Some phases sequential (plan→implement), some parallel (review+test). +- **This is actually what we're proposing** - Respects natural dependencies while maximizing parallelism. + +## Success Metrics + +1. **Time to PR** - Should be faster than sequential `/lfg` for non-trivial features +2. **Context efficiency** - No single agent exceeds context limits +3. **Completion rate** - High percentage of swarms complete successfully +4. **User satisfaction** - Clear progress visibility, fewer interruptions + +--- + +**Next Steps:** +1. Review and approve this plan +2. Implement MVP (Phase 1) +3. Test with a real feature implementation +4. Iterate based on learnings diff --git a/plugins/compound-engineering/.claude-plugin/plugin.json b/plugins/compound-engineering/.claude-plugin/plugin.json index 97ea742..a9f29f9 100644 --- a/plugins/compound-engineering/.claude-plugin/plugin.json +++ b/plugins/compound-engineering/.claude-plugin/plugin.json @@ -1,7 +1,7 @@ { "name": "compound-engineering", - "version": "2.28.0", - "description": "AI-powered development tools. 28 agents, 24 commands, 15 skills, 1 MCP server for code review, research, design, and workflow automation.", + "version": "2.29.0", + "description": "AI-powered development tools. 28 agents, 25 commands, 15 skills, 1 MCP server for code review, research, design, and workflow automation.", "author": { "name": "Kieran Klaassen", "email": "kieran@every.to", diff --git a/plugins/compound-engineering/CHANGELOG.md b/plugins/compound-engineering/CHANGELOG.md index dd1c7f9..3f9eb59 100644 --- a/plugins/compound-engineering/CHANGELOG.md +++ b/plugins/compound-engineering/CHANGELOG.md @@ -5,6 +5,26 @@ All notable changes to the compound-engineering plugin will be documented in thi The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). +## [2.29.0] - 2026-01-25 + +### Added + +- **`/slfg` command** - Swarm-based autonomous engineering workflow using TeammateTool + - Spawns specialized agents (Planner, Researcher, Worker, Reviewer, Tester, Resolver, Video) that work in parallel + - Coordinates through shared task list with dependencies + - Team lead orchestrates and approves plans before implementation + - Maximizes parallelism: planning + research run simultaneously, review + testing run simultaneously + - Full autonomous workflow from feature description to PR with video + +--- + +## [2.28.1] - 2026-01-22 + +### Fixed + +- **Documentation** - Updated component counts and added missing entries in README, HTML docs, and metadata to reflect 28 agents, 24 commands, and 15 skills +- **Slash Commands** - Corrected documentation for recently added commands (`lfg`, `release-docs`, `deploy-docs`) + ## [2.28.0] - 2026-01-21 ### Added diff --git a/plugins/compound-engineering/README.md b/plugins/compound-engineering/README.md index b1a710d..78ef65e 100644 --- a/plugins/compound-engineering/README.md +++ b/plugins/compound-engineering/README.md @@ -6,9 +6,9 @@ AI-powered development tools that get smarter with every use. Make each unit of | Component | Count | |-----------|-------| -| Agents | 27 | -| Commands | 20 | -| Skills | 14 | +| Agents | 28 | +| Commands | 25 | +| Skills | 15 | | MCP Servers | 1 | ## Agents @@ -34,13 +34,14 @@ Agents are organized into categories for easier discovery. | `security-sentinel` | Security audits and vulnerability assessments | | `julik-frontend-races-reviewer` | Review JavaScript/Stimulus code for race conditions | -### Research (4) +### Research (5) | Agent | Description | |-------|-------------| | `best-practices-researcher` | Gather external best practices and examples | | `framework-docs-researcher` | Research framework documentation and best practices | | `git-history-analyzer` | Analyze git history and code evolution | +| `learnings-researcher` | Research and synthesize team learnings and patterns | | `repo-research-analyst` | Research repository structure and conventions | ### Design (3) @@ -85,12 +86,17 @@ Core workflow commands use `workflows:` prefix to avoid collisions with built-in | Command | Description | |---------|-------------| +| `/agent-native-audit` | Audit features for agent-native compliance | | `/deepen-plan` | Enhance plans with parallel research agents for each section | | `/changelog` | Create engaging changelogs for recent merges | | `/create-agent-skill` | Create or edit Claude Code skills | +| `/deploy-docs` | Deploy documentation to GitHub Pages | | `/generate_command` | Generate new slash commands | | `/heal-skill` | Fix skill documentation issues | +| `/lfg` | Sequential autonomous engineering workflow | +| `/slfg` | Swarm-based autonomous engineering with parallel agents | | `/plan_review` | Multi-agent plan review in parallel | +| `/release-docs` | Build and release plugin documentation | | `/report-bug` | Report a bug in the plugin | | `/reproduce-bug` | Reproduce bugs using logs and console | | `/resolve_parallel` | Resolve TODO comments in parallel | @@ -108,6 +114,7 @@ Core workflow commands use `workflows:` prefix to avoid collisions with built-in | Skill | Description | |-------|-------------| | `agent-native-architecture` | Build AI agents using prompt-native architecture | +| `brainstorming` | Expert guidance for brainstorming and exploring ideas | ### Development Tools diff --git a/plugins/compound-engineering/commands/slfg.md b/plugins/compound-engineering/commands/slfg.md new file mode 100644 index 0000000..b9fb506 --- /dev/null +++ b/plugins/compound-engineering/commands/slfg.md @@ -0,0 +1,557 @@ +--- +name: slfg +description: Swarm-based autonomous engineering workflow with parallel agents +argument-hint: "[feature description]" +--- + +# Swarm LFG - Full Autonomous Engineering with Parallel Agents + +Transform a feature description into a shipped PR using a coordinated swarm of specialized agents working in parallel. + +## Feature Description + + #$ARGUMENTS + +**If the feature description above is empty, ask the user:** "What feature would you like to build? Please describe it clearly." + +Do not proceed until you have a clear feature description. + +## Swarm Architecture + +``` + ┌─────────────────┐ + │ TEAM LEAD │ + │ (You - Orchestrator) │ + └────────┬────────┘ + │ + ┌───────────────────┼───────────────────┐ + ▼ ▼ ▼ + ┌──────────┐ ┌──────────┐ ┌──────────┐ + │ Planner │ │Researcher│ │ Worker │ + └──────────┘ └──────────┘ └──────────┘ + │ │ │ + ▼ ▼ ▼ + ┌──────────┐ ┌──────────┐ ┌──────────┐ + │ Reviewer │ │ Tester │ │Resolver(s)│ + └──────────┘ └──────────┘ └──────────┘ + │ + ▼ + ┌──────────┐ + │ Video │ + └──────────┘ +``` + +## Step 1: Create Swarm Team + +Create a new team for this engineering swarm: + +``` +TeammateTool: + operation: spawnTeam + team_name: "slfg-{YYYYMMDD-HHMMSS}" + description: "Swarm for: {feature_description}" +``` + +Store the team name for use throughout this workflow. + +## Step 2: Create Task List with Dependencies + +Create all tasks upfront with their dependency relationships: + +### Task 1: Create Implementation Plan +``` +TaskCreate: + subject: "Create implementation plan" + description: | + Run /workflows:plan for: {feature_description} + Write plan to docs/plans/ following naming convention. + Use ExitPlanMode when ready for approval. + activeForm: "Planning feature" +``` + +### Task 2: Research Best Practices (Parallel with Task 1) +``` +TaskCreate: + subject: "Research best practices and patterns" + description: | + Research using: + - Context7 for framework documentation + - WebSearch for current best practices + - Codebase patterns via repo-research-analyst + Write findings to enhance the plan. + activeForm: "Researching patterns" +``` + +### Task 3: Implement Feature (Blocked by 1 & 2) +``` +TaskCreate: + subject: "Implement the approved plan" + description: | + Run /workflows:work on the approved plan. + Create incremental commits. + Push to feature branch. + activeForm: "Implementing feature" +``` +Then: `TaskUpdate taskId: 3, addBlockedBy: ["1", "2"]` + +### Task 4: Review Implementation (Blocked by 3) +``` +TaskCreate: + subject: "Review code changes" + description: | + Run parallel review agents: + - kieran-rails-reviewer + - dhh-rails-reviewer + - code-simplicity-reviewer + - security-sentinel + - performance-oracle + Create todos for findings in todos/ directory. + activeForm: "Reviewing implementation" +``` +Then: `TaskUpdate taskId: 4, addBlockedBy: ["3"]` + +### Task 5: Run Browser Tests (Blocked by 3, Parallel with 4) +``` +TaskCreate: + subject: "Run browser tests" + description: | + Run /test-browser on affected pages. + Create P1 todos for any failures. + activeForm: "Running browser tests" +``` +Then: `TaskUpdate taskId: 5, addBlockedBy: ["3"]` + +### Task 6: Resolve Findings (Blocked by 4 & 5) +``` +TaskCreate: + subject: "Fix review findings and test failures" + description: | + Run /resolve_todo_parallel on all pending todos. + Re-run tests after fixes. + activeForm: "Resolving findings" +``` +Then: `TaskUpdate taskId: 6, addBlockedBy: ["4", "5"]` + +### Task 7: Record Feature Video (Blocked by 6) +``` +TaskCreate: + subject: "Record feature demo video" + description: | + Run /feature-video for the PR. + Upload and embed in PR description. + activeForm: "Recording demo video" +``` +Then: `TaskUpdate taskId: 7, addBlockedBy: ["6"]` + +### Task 8: Create Pull Request (Blocked by 7) +``` +TaskCreate: + subject: "Create pull request" + description: | + Push all changes. + Create PR with summary, screenshots, video. + Include Compound Engineered badge. + activeForm: "Creating pull request" +``` +Then: `TaskUpdate taskId: 8, addBlockedBy: ["7"]` + +## Step 3: Spawn Phase 1 Agents (Parallel) + +Spawn Planner and Researcher agents to work simultaneously: + +### Spawn Planner Agent +``` +Task: + subagent_type: general-purpose + team_name: "{team_name}" + name: "planner" + mode: "plan" + prompt: | + # Planner Agent + + You are the **Planner** for the slfg engineering swarm. + + ## Your Task + TaskGet taskId: "1" to read your full assignment. + + ## Instructions + 1. Claim your task: TaskUpdate taskId: "1", status: "in_progress", owner: "planner" + 2. Run /workflows:plan for: {feature_description} + 3. Create a comprehensive plan in docs/plans/ + 4. When plan is ready, use ExitPlanMode to request team lead approval + 5. After approval, mark complete: TaskUpdate taskId: "1", status: "completed" + + ## Communication + - You are part of team: {team_name} + - Your name is: planner + - Send messages to team-lead if you have questions or are blocked + + Work autonomously. Your plan will be reviewed before implementation begins. +``` + +### Spawn Researcher Agent +``` +Task: + subagent_type: general-purpose + team_name: "{team_name}" + name: "researcher" + prompt: | + # Researcher Agent + + You are the **Researcher** for the slfg engineering swarm. + + ## Your Task + TaskGet taskId: "2" to read your full assignment. + + ## Instructions + 1. Claim your task: TaskUpdate taskId: "2", status: "in_progress", owner: "researcher" + 2. Research best practices for: {feature_description} + 3. Use these tools: + - mcp__plugin_compound-engineering_context7__resolve-library-id + - mcp__plugin_compound-engineering_context7__query-docs + - WebSearch for current (2026) best practices + - Grep/Glob to find similar patterns in codebase + 4. Write findings to docs/research/ or add to the plan file + 5. Mark complete: TaskUpdate taskId: "2", status: "completed" + + ## Communication + - You are part of team: {team_name} + - Your name is: researcher + - Send findings to planner if they should inform the plan + - Send messages to team-lead if blocked + + Work autonomously and thoroughly. +``` + +**Launch both agents in parallel** (single message with multiple Task tool calls). + +## Step 4: Handle Plan Approval + +As team lead, you will receive a `plan_approval_request` message from the Planner agent. + +### When you receive the plan approval request: + +1. **Read the plan file** to review what's proposed +2. **Evaluate the plan:** + - Is it complete and actionable? + - Does it follow project conventions? + - Are dependencies identified? + - Is the scope appropriate? + +3. **Approve or Reject:** + + **To Approve:** + ``` + TeammateTool: + operation: approvePlan + target_agent_id: "planner" + request_id: "{requestId from message}" + ``` + + **To Reject with Feedback:** + ``` + TeammateTool: + operation: rejectPlan + target_agent_id: "planner" + request_id: "{requestId from message}" + feedback: "Please add more detail about error handling..." + ``` + +4. **Wait for both Phase 1 tasks to complete** before proceeding. + +## Step 5: Spawn Worker Agent + +Once Tasks 1 and 2 are complete: + +``` +Task: + subagent_type: general-purpose + team_name: "{team_name}" + name: "worker" + prompt: | + # Worker Agent + + You are the **Worker** for the slfg engineering swarm. + + ## Your Task + TaskGet taskId: "3" to read your full assignment. + + ## Instructions + 1. Claim your task: TaskUpdate taskId: "3", status: "in_progress", owner: "worker" + 2. Read the approved plan at docs/plans/{plan_file} + 3. Run /workflows:work following the plan exactly + 4. Create incremental commits with conventional messages + 5. Push to feature branch (never main) + 6. Mark complete: TaskUpdate taskId: "3", status: "completed" + + ## Communication + - You are part of team: {team_name} + - Your name is: worker + - Send progress updates to team-lead for significant milestones + - Send messages to team-lead if blocked + + Work autonomously. Focus on completing the implementation. +``` + +## Step 6: Spawn Reviewer and Tester (Parallel) + +Once Task 3 (implementation) is complete, spawn both: + +### Spawn Reviewer Agent +``` +Task: + subagent_type: general-purpose + team_name: "{team_name}" + name: "reviewer" + prompt: | + # Reviewer Agent + + You are the **Reviewer** for the slfg engineering swarm. + + ## Your Task + TaskGet taskId: "4" to read your full assignment. + + ## Instructions + 1. Claim your task: TaskUpdate taskId: "4", status: "in_progress", owner: "reviewer" + 2. Get the changed files: git diff main...HEAD --name-only + 3. Spawn parallel review agents: + - Task kieran-rails-reviewer("Review the changes") + - Task code-simplicity-reviewer("Review for simplicity") + - Task security-sentinel("Check for security issues") + - Task performance-oracle("Check for performance issues") + 4. Create todos for ALL findings in todos/ directory + 5. Mark complete: TaskUpdate taskId: "4", status: "completed" + + ## Communication + - You are part of team: {team_name} + - Your name is: reviewer + - Send summary to team-lead when complete + + Document findings but do NOT fix them. Resolvers handle fixes. +``` + +### Spawn Tester Agent +``` +Task: + subagent_type: general-purpose + team_name: "{team_name}" + name: "tester" + prompt: | + # Tester Agent + + You are the **Tester** for the slfg engineering swarm. + + ## Your Task + TaskGet taskId: "5" to read your full assignment. + + ## Instructions + 1. Claim your task: TaskUpdate taskId: "5", status: "in_progress", owner: "tester" + 2. Identify affected pages from git diff + 3. Run /test-browser on each affected page + 4. Create P1 todos for any test failures in todos/ + 5. Mark complete: TaskUpdate taskId: "5", status: "completed" + + ## Communication + - You are part of team: {team_name} + - Your name is: tester + - Send results summary to team-lead + + Document failures as todos but do NOT fix them. +``` + +**Launch both in parallel.** + +## Step 7: Spawn Resolver Agents + +Once Tasks 4 and 5 are complete: + +``` +Task: + subagent_type: general-purpose + team_name: "{team_name}" + name: "resolver" + prompt: | + # Resolver Agent + + You are the **Resolver** for the slfg engineering swarm. + + ## Your Task + TaskGet taskId: "6" to read your full assignment. + + ## Instructions + 1. Claim your task: TaskUpdate taskId: "6", status: "in_progress", owner: "resolver" + 2. List all pending todos: ls todos/*-pending-*.md + 3. For each todo, spawn a parallel resolver: + - Task pr-comment-resolver("Fix: {todo_description}") + 4. Re-run tests after fixes if needed + 5. Mark complete: TaskUpdate taskId: "6", status: "completed" + + ## Communication + - You are part of team: {team_name} + - Your name is: resolver + + Fix all findings. Commit after each fix. +``` + +## Step 8: Spawn Video Agent + +Once Task 6 is complete: + +``` +Task: + subagent_type: general-purpose + team_name: "{team_name}" + name: "video" + prompt: | + # Video Agent + + You are the **Video** agent for the slfg engineering swarm. + + ## Your Task + TaskGet taskId: "7" to read your full assignment. + + ## Instructions + 1. Claim your task: TaskUpdate taskId: "7", status: "in_progress", owner: "video" + 2. Run /feature-video for the current PR + 3. Ensure video is uploaded and embedded in PR + 4. Mark complete: TaskUpdate taskId: "7", status: "completed" + + ## Communication + - You are part of team: {team_name} + - Your name is: video +``` + +## Step 9: Create Pull Request + +Once Task 7 is complete, handle Task 8 yourself (Team Lead): + +1. Claim the task: `TaskUpdate taskId: "8", status: "in_progress", owner: "team-lead"` + +2. Push all changes: + ```bash + git push -u origin {branch_name} + ``` + +3. Create PR: + ```bash + gh pr create --title "feat: {feature_description}" --body "$(cat <<'EOF' + ## Summary + - {what was built} + - {key decisions} + + ## Demo + {video embed from /feature-video} + + ## Testing + - Browser tests: {pass/fail} + - Review agents: {findings resolved} + + ## Swarm Stats + - Agents spawned: {count} + - Tasks completed: 8/8 + - Team: {team_name} + + --- + [![Compound Engineered](https://img.shields.io/badge/Compound-Engineered-6366f1)](https://github.com/EveryInc/compound-engineering-plugin) + Autonomous swarm with [Claude Code](https://claude.com/claude-code) + EOF + )" + ``` + +4. Mark complete: `TaskUpdate taskId: "8", status: "completed"` + +## Step 10: Cleanup and Report + +### Shutdown Teammates + +Send shutdown requests to all remaining teammates: + +``` +For each active teammate: + TeammateTool: + operation: requestShutdown + target_agent_id: "{teammate_name}" + reason: "Swarm complete - PR created" +``` + +### Cleanup Team Resources + +``` +TeammateTool: + operation: cleanup +``` + +### Final Report + +Present the completion summary: + +```markdown +## Swarm Complete! + +**Feature:** {feature_description} +**PR:** {pr_url} +**Branch:** {branch_name} +**Team:** {team_name} + +### Task Summary +| Task | Agent | Status | +|------|-------|--------| +| Create plan | planner | Complete | +| Research | researcher | Complete | +| Implement | worker | Complete | +| Review | reviewer | Complete | +| Test | tester | Complete | +| Resolve | resolver | Complete | +| Video | video | Complete | +| Create PR | team-lead | Complete | + +### Artifacts +- Plan: docs/plans/{plan_file} +- PR: {pr_url} +- Video: {video_url} + +### Swarm Stats +- Total agents spawned: {count} +- Messages exchanged: {count} +- Time elapsed: {duration} +``` + +## Error Handling + +### If an agent fails: +1. Check the error message +2. Option to respawn the agent with the same task +3. Or handle manually and mark task complete + +### If plan is rejected: +1. Planner receives feedback and revises +2. Planner resubmits with ExitPlanMode +3. Repeat until approved + +### If tests fail: +1. Tester creates P1 todos +2. Resolver fixes issues +3. Tester re-runs tests +4. Continue when all pass + +### If user interrupts: +1. Cleanup team resources +2. Preserve work on branch +3. Can resume later with new swarm + +## Key Principles + +1. **Parallelism** - Run independent phases simultaneously +2. **Specialization** - Each agent focuses on one role +3. **Autonomy** - Agents work independently, coordinate through tasks +4. **Visibility** - Team lead maintains oversight of all progress +5. **Resilience** - Failures are isolated, recoverable + +## When to Use /slfg vs /lfg + +| Use /slfg when... | Use /lfg when... | +|-------------------|------------------| +| Feature is complex (multiple phases) | Simple, linear workflow | +| Want maximum parallelism | Prefer sequential execution | +| Multiple independent workstreams | Single focused task | +| Want to see swarm coordination | Want simpler mental model | From 0b651c6bfeea1d7fb8a68a45f578fb2e1dfc1c87 Mon Sep 17 00:00:00 2001 From: Kieran Klaassen Date: Sun, 25 Jan 2026 19:44:10 -0500 Subject: [PATCH 2/3] feat(commands): add /swarm-status command to show active swarm teams Add new /swarm-status command that displays: - Active teams from ~/.claude/teams/ - Teammates with running/stopped status - Task progress with completion percentage Simplified after code review to focus on essential information. Built using the /slfg swarm workflow with parallel agents. Co-Authored-By: Claude Opus 4.5 --- .../.claude-plugin/plugin.json | 4 +- plugins/compound-engineering/CHANGELOG.md | 12 ++++ plugins/compound-engineering/README.md | 3 +- .../commands/swarm-status.md | 63 +++++++++++++++++++ 4 files changed, 79 insertions(+), 3 deletions(-) create mode 100644 plugins/compound-engineering/commands/swarm-status.md diff --git a/plugins/compound-engineering/.claude-plugin/plugin.json b/plugins/compound-engineering/.claude-plugin/plugin.json index a9f29f9..725fe0b 100644 --- a/plugins/compound-engineering/.claude-plugin/plugin.json +++ b/plugins/compound-engineering/.claude-plugin/plugin.json @@ -1,7 +1,7 @@ { "name": "compound-engineering", - "version": "2.29.0", - "description": "AI-powered development tools. 28 agents, 25 commands, 15 skills, 1 MCP server for code review, research, design, and workflow automation.", + "version": "2.30.0", + "description": "AI-powered development tools. 28 agents, 26 commands, 15 skills, 1 MCP server for code review, research, design, and workflow automation.", "author": { "name": "Kieran Klaassen", "email": "kieran@every.to", diff --git a/plugins/compound-engineering/CHANGELOG.md b/plugins/compound-engineering/CHANGELOG.md index 3f9eb59..ede43ba 100644 --- a/plugins/compound-engineering/CHANGELOG.md +++ b/plugins/compound-engineering/CHANGELOG.md @@ -5,6 +5,18 @@ All notable changes to the compound-engineering plugin will be documented in thi The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). +## [2.30.0] - 2026-01-25 + +### Added + +- **`/swarm-status` command** - Display status of active swarm teams + - Shows all active teams from ~/.claude/teams/ + - Lists teammates with their roles and status (Active/Idle/Offline) + - Displays task progress with completion percentage + - Supports viewing specific team or all teams + +--- + ## [2.29.0] - 2026-01-25 ### Added diff --git a/plugins/compound-engineering/README.md b/plugins/compound-engineering/README.md index 78ef65e..726c388 100644 --- a/plugins/compound-engineering/README.md +++ b/plugins/compound-engineering/README.md @@ -7,7 +7,7 @@ AI-powered development tools that get smarter with every use. Make each unit of | Component | Count | |-----------|-------| | Agents | 28 | -| Commands | 25 | +| Commands | 26 | | Skills | 15 | | MCP Servers | 1 | @@ -95,6 +95,7 @@ Core workflow commands use `workflows:` prefix to avoid collisions with built-in | `/heal-skill` | Fix skill documentation issues | | `/lfg` | Sequential autonomous engineering workflow | | `/slfg` | Swarm-based autonomous engineering with parallel agents | +| `/swarm-status` | Show status of active swarm teams and tasks | | `/plan_review` | Multi-agent plan review in parallel | | `/release-docs` | Build and release plugin documentation | | `/report-bug` | Report a bug in the plugin | diff --git a/plugins/compound-engineering/commands/swarm-status.md b/plugins/compound-engineering/commands/swarm-status.md new file mode 100644 index 0000000..9240cc8 --- /dev/null +++ b/plugins/compound-engineering/commands/swarm-status.md @@ -0,0 +1,63 @@ +--- +name: swarm-status +description: Show status of active swarm teams, teammates, and tasks +argument-hint: "[team-name or 'all']" +--- + +# Swarm Status + +Display the current status of active swarm teams. + +## Usage + +```bash +/swarm-status # Show all active teams +/swarm-status my-team # Show specific team +``` + +## Step 1: Find Active Teams + +```bash +ls ~/.claude/teams/*/config.json 2>/dev/null +``` + +If no teams found: +``` +No active swarm teams found. + +Start a new swarm with: /slfg "your feature description" +``` + +## Step 2: Display Team Status + +For each team, read config and display: + +```markdown +## Swarm: {team_name} + +**Description:** {description} +**Lead:** {lead_agent_id} + +### Teammates ({count}) + +| Name | Status | +|------|--------| +| team-lead | Running | +| planner | Running | +| worker | Stopped | + +### Tasks + +| # | Status | Subject | Owner | +|---|--------|---------|-------| +| 1 | Complete | Create plan | planner | +| 2 | In Progress | Implement | worker | +| 3 | Pending | Review | - | + +**Progress:** 1/3 complete (33%) +``` + +## Status Detection + +- **Running** - Agent has `backendType: in-process` or active `tmuxPaneId` +- **Stopped** - Otherwise From be0fff61f63313f70cb6712555724b8b661f5519 Mon Sep 17 00:00:00 2001 From: Kieran Klaassen Date: Sun, 25 Jan 2026 22:53:05 -0500 Subject: [PATCH 3/3] chore: update docs and marketplace for v2.30.0 - Update marketplace.json version and command count - Update HTML docs with new commands (/slfg, /swarm-status) - Add swarm-status plan document Co-Authored-By: Claude Opus 4.5 --- .claude-plugin/marketplace.json | 4 +- docs/index.html | 78 +++++++++++--- docs/pages/agents.html | 89 +++++++++++---- docs/pages/commands.html | 89 ++++++++++++--- docs/pages/skills.html | 75 +++++++++---- ...26-01-25-feat-swarm-status-command-plan.md | 101 ++++++++++++++++++ 6 files changed, 364 insertions(+), 72 deletions(-) create mode 100644 docs/plans/2026-01-25-feat-swarm-status-command-plan.md diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.json index 4e22718..bf45358 100644 --- a/.claude-plugin/marketplace.json +++ b/.claude-plugin/marketplace.json @@ -11,8 +11,8 @@ "plugins": [ { "name": "compound-engineering", - "description": "AI-powered development tools that get smarter with every use. Make each unit of engineering work easier than the last. Includes 28 specialized agents, 24 commands, and 15 skills.", - "version": "2.28.0", + "description": "AI-powered development tools that get smarter with every use. Make each unit of engineering work easier than the last. Includes 28 specialized agents, 26 commands, and 15 skills.", + "version": "2.30.0", "author": { "name": "Kieran Klaassen", "url": "https://github.com/kieranklaassen", diff --git a/docs/index.html b/docs/index.html index ec26ee3..28f4484 100644 --- a/docs/index.html +++ b/docs/index.html @@ -244,7 +244,7 @@

Delegate

The security-sentinel has checked 10,000 PRs for SQL injection. The kieran-rails-reviewer never approves a controller with business logic. They don't get tired, don't skip Friday afternoon reviews, don't forget the conventions you agreed on in March. Run /work and watch your plan execute with quality gates that actually enforce your standards—every single time.

- 27 specialized agents + 28 specialized agents /work dhh-rails-style skill git-worktree skill @@ -292,7 +292,7 @@

Compound

- 27 Specialized Agents + 28 Specialized Agents

Think of them as coworkers who never quit. The security-sentinel has seen every SQL injection variant. The kieran-rails-reviewer enforces conventions with zero compromise. The performance-oracle spots N+1 queries while you're still reading the PR. Run them solo or launch twelve in parallel—your choice. @@ -301,7 +301,7 @@

-

Review Agents (11)

+

Review Agents (14)

@@ -396,7 +396,7 @@

Review Agents (11)

-

Research Agents (4)

+

Research Agents (5)

@@ -430,6 +430,14 @@

Research Agents (4)

Research repository structure and conventions. Understand project patterns and organization.

claude agent repo-research-analyst
+
+
+ learnings-researcher + Research +
+

Research and synthesize team learnings and patterns from documentation and history.

+ claude agent learnings-researcher +
@@ -531,7 +539,7 @@

Documentation Agent (1)

- 19 Powerful Commands + 24 Powerful Commands

Slash commands that replace entire workflows. /review is your code review committee. /plan is your research team. /triage sorts 50 todos in the time it takes you to read five. Each one automates hours of work into a single line. @@ -540,7 +548,7 @@

-

Workflow Commands

+

Workflow Commands (5)

@@ -570,12 +578,19 @@

Workflow Commands

Document solved problems to compound team knowledge. Turn learnings into reusable patterns.

+
+
+ /brainstorm + core +
+

Explore requirements and approaches before planning. Brainstorm with multiple experts.

+
-

Utility Commands

+

Utility Commands (19)

@@ -612,13 +627,6 @@

Utility Commands

Multi-agent plan review in parallel.

-
-
- /prime - util -
-

Prime/setup command for project initialization.

-
/report-bug @@ -675,6 +683,20 @@

Utility Commands

Validate and prepare documentation for GitHub Pages deployment.

+
+
+ /lfg + util +
+

Launch something fun (or just start the day).

+
+
+
+ /agent-native-audit + util +
+

Audit features for agent-native compliance.

+

@@ -683,7 +705,7 @@

Utility Commands

- 12 Intelligent Skills + 15 Intelligent Skills

Domain expertise on tap. Need to write a Ruby gem? The andrew-kane-gem-writer knows the patterns Andrew uses in 50+ popular gems. Building a Rails app? The dhh-rails-style enforces 37signals conventions. Generating images? The gemini-imagegen has Google's AI on speed dial. Just invoke the skill and watch it work. @@ -750,6 +772,14 @@

Development Tools

Capture solved problems as categorized documentation with YAML schema.

skill: compound-docs
+
+
+ brainstorming + Ideation +
+

Expert guidance for brainstorming and exploring ideas with multiple agent perspectives.

+ skill: brainstorming +
@@ -781,6 +811,22 @@

Content & Workflow

Manage Git worktrees for parallel development on multiple branches.

skill: git-worktree
+
+
+ rclone + Cloud +
+

Upload and sync files to S3, Cloudflare R2, and other cloud providers.

+ skill: rclone +
+
+
+ agent-browser + Automation +
+

Browser automation and interaction using agent-browser CLI.

+ skill: agent-browser +
@@ -989,7 +1035,7 @@

Frequently Asked Questions

Free & Open Source

Install Once. Compound Forever.

- Your next code review takes 30 seconds. The one after that? Even faster. That's compounding. Get 27 expert agents, 19 workflow commands, and 12 specialized skills working for you right now. + Your next code review takes 30 seconds. The one after that? Even faster. That's compounding. Get 28 expert agents, 24 workflow commands, and 15 specialized skills working for you right now.

diff --git a/docs/pages/agents.html b/docs/pages/agents.html index eb39b75..2952580 100644 --- a/docs/pages/agents.html +++ b/docs/pages/agents.html @@ -4,7 +4,7 @@ Agent Reference - Compounding Engineering - + @@ -33,9 +33,9 @@

Getting Started

Reference
@@ -48,8 +48,8 @@

Resources

On This Page

    -
  • Review (10)
    • -
  • Research (4)
    • +
  • Review (14)
    • +
  • Research (5)
  • Workflow (5)
  • Design (3)
  • Docs (1)
    • @@ -75,7 +75,7 @@

    On This Page

    Agent Reference

    - Think of agents as your expert teammates who never sleep. You've got 23 specialists here—each one obsessed with a single domain. Call them individually when you need focused expertise, or orchestrate them together for multi-angle analysis. They're opinionated, they're fast, and they remember your codebase better than you do. + Think of agents as your expert teammates who never sleep. You've got 28 specialists here—each one obsessed with a single domain. Call them individually when you need focused expertise, or orchestrate them together for multi-angle analysis. They're opinionated, they're fast, and they remember your codebase better than you do.

    @@ -95,7 +95,7 @@

    How to Use Agents

    -

    Review Agents (10)

    +

    Review Agents (14)

    Your code review dream team. These agents catch what humans miss at 2am—security holes, performance cliffs, architectural drift, and those "it works but I hate it" moments. They're picky. They disagree with each other. That's the point.

    @@ -295,32 +295,62 @@

    Detection Areas

    -
    +
    -

    code-simplicity-reviewer

    - Quality +

    julik-frontend-races-reviewer

    + JavaScript

    - Simplicity is violent discipline. This agent asks "do you actually need this?" about every line, every abstraction, every dependency. YAGNI isn't a suggestion—it's the law. Your 200-line feature with three layers of indirection? This agent will show you the 50-line version that does the same thing. Complexity is a liability; simplicity compounds. + Review JavaScript and Stimulus code for race conditions, DOM event handling, promise management, and timer cleanup. Spotting UI data races before they cause "it only happens sometimes" bugs.

    -

    Simplification Checks

    -
      -
    • Analyze every line for necessity
      • -
    • Simplify complex logic
      • -
    • Remove redundancy and duplication
      • -
    • Challenge abstractions
      • -
    • Optimize for readability
      • -
    • Eliminate premature generalization
      • -
    - 
    claude agent code-simplicity-reviewer
    + 
    claude agent julik-frontend-races-reviewer
    +
    +
    + +
    +
    +

    agent-native-reviewer

    + AI +
    +

    + Verify features are agent-native. Ensures action and context parity—anything a user can do or see, an agent must also be able to do or see. +

    +
    + 
    claude agent agent-native-reviewer
    +
    +
    + +
    +
    +

    data-migration-expert

    + Data +
    +

    + Validate ID mappings match production reality. Check for swapped values and verify rollback safety during schema changes and data transformations. +

    +
    + 
    claude agent data-migration-expert
    +
    +
    + +
    +
    +

    deployment-verification-agent

    + Ops +
    +

    + Create Go/No-Go deployment checklists for risky data changes. Essential for PRs touching production data, migrations, or critical backfills. +

    +
    + 
    claude agent deployment-verification-agent
-

Research Agents (4)

+

Research Agents (5)

Stop guessing. These agents dig through documentation, GitHub repos, git history, and real-world examples to give you answers backed by evidence. They read faster than you, remember more than you, and synthesize patterns you'd miss. Perfect for "how should I actually do this?" questions.

@@ -406,6 +436,19 @@

Analysis Areas

claude agent repo-research-analyst
+ +
+
+

learnings-researcher

+ Research +
+

+ Research and synthesize team learnings and patterns from documentation, history, and solved problems. Helps turn individual fixes into collective knowledge. +

+
+ 
claude agent learnings-researcher
+
+
diff --git a/docs/pages/commands.html b/docs/pages/commands.html index c5be692..b416dab 100644 --- a/docs/pages/commands.html +++ b/docs/pages/commands.html @@ -4,7 +4,7 @@ Command Reference - Compounding Engineering - + @@ -33,9 +33,9 @@

Getting Started

Reference

@@ -48,8 +48,8 @@

Resources

On This Page

@@ -77,8 +77,23 @@

Command Reference

-

Workflow Commands (four)

-

These are the big four: Plan your feature, Review your code, Work through the implementation, and Codify what you learned. Every professional developer does this cycle—these commands just make you faster at it.

+

Workflow Commands (5)

+

These are the core five: Brainstorm your ideas, Plan your feature, Review your code, Work through the implementation, and Codify what you learned. Every professional developer does this cycle—these commands just make you faster at it.

+ +
+
+ /brainstorm +
+

+ Explore requirements and approaches before planning. Brainstorm with multiple experts to find the best architectural path forward. +

+

Arguments

+

[topic or problem description]

+
+ 
/brainstorm architecture for new notification system
+/brainstorm how to optimize database performance
+
+
@@ -244,7 +259,7 @@

Auto-Triggers

-

Utility Commands (12)

+

Utility Commands (19)

The supporting cast—commands that do one specific thing really well. Generate changelogs, resolve todos in parallel, triage findings, create new commands. The utilities you reach for daily.

@@ -485,15 +500,63 @@

Process

-
+
+
+ /agent-native-audit +
+

+ Audit features for agent-native compliance. Ensures that all functionality available to users is also accessible to AI agents. +

+
+ 
/agent-native-audit
+
+
+ +
+
+ /release-docs +
+

+ Build and update the documentation site with current plugin components. Synchronizes all reference pages with the current state of the plugin. +

+
+ 
/release-docs
+
+
+ +
+
+ /deploy-docs +
+

+ Validate and prepare documentation for GitHub Pages deployment. Ensures all internal links work and assets are correctly placed. +

+
+ 
/deploy-docs
+
+
+ +
+
+ /lfg +
+

+ Launch something fun (or just start the day). A mood-boosting command to kick off your development session. +

+
+ 
/lfg
+
+
+ +
- /prime + /deepen-plan

- Your project initialization command. What exactly it does depends on your project setup—think of it as the "get everything ready" button before you start coding. + Enhance plans with parallel research agents for each section. Adds more depth and research to an existing implementation plan.

- 
/prime
+ 
/deepen-plan plans/my-plan.md
diff --git a/docs/pages/skills.html b/docs/pages/skills.html index a86ae91..58b1449 100644 --- a/docs/pages/skills.html +++ b/docs/pages/skills.html @@ -4,7 +4,7 @@ Skill Reference - Compounding Engineering - + @@ -33,9 +33,9 @@

Getting Started

Reference

@@ -48,8 +48,8 @@

Resources

On This Page

@@ -101,7 +101,7 @@

Skills vs Agents

-

Development Tools (8)

+

Development Tools (10)

These skills teach Claude specific coding styles and architectural patterns. Use them when you want code that follows a particular philosophy—not just any working code, but code that looks like it was written by a specific person or framework.

@@ -356,11 +356,24 @@

Core Principle

skill: agent-native-architecture
+ +
+
+

brainstorming

+ Ideation +
+

+ Expert guidance for brainstorming and exploring ideas. Uses multiple agent perspectives (Creative, Skeptic, Pragmatist) to stress-test ideas and find the best path forward. +

+
+ 
skill: brainstorming
+
+
-

Content & Workflow (3)

+

Content & Workflow (4)

Writing, editing, and organizing work. These skills handle everything from style guide compliance to git worktree management—the meta-work that makes the real work easier.

@@ -447,16 +460,16 @@

git-worktree

Commands

# Create new worktree
-bash scripts/worktree-manager.sh create feature-login
-
-# List worktrees
-bash scripts/worktree-manager.sh list
-
-# Switch to worktree
-bash scripts/worktree-manager.sh switch feature-login
-
-# Clean up completed worktrees
-bash scripts/worktree-manager.sh cleanup
+00450| bash scripts/worktree-manager.sh create feature-login +00451| +00452| # List worktrees +00453| bash scripts/worktree-manager.sh list +00454| +00455| # Switch to worktree +00456| bash scripts/worktree-manager.sh switch feature-login +00457| +00458| # Clean up completed worktrees +00459| bash scripts/worktree-manager.sh cleanup

Integration

    @@ -472,6 +485,32 @@

    Requirements

    skill: git-worktree
+ +
+
+

rclone

+ Cloud +
+

+ Upload, sync, and manage files across cloud storage providers using rclone. Supports S3, Cloudflare R2, Backblaze B2, Google Drive, and more. +

+
+ 
skill: rclone
+
+
+ +
+
+

agent-browser

+ Automation +
+

+ Browser automation using Vercel's agent-browser CLI. Interact with web pages, fill forms, take screenshots, and scrape data through AI-driven automation. +

+
+ 
skill: agent-browser
+
+
diff --git a/docs/plans/2026-01-25-feat-swarm-status-command-plan.md b/docs/plans/2026-01-25-feat-swarm-status-command-plan.md new file mode 100644 index 0000000..e3b74f8 --- /dev/null +++ b/docs/plans/2026-01-25-feat-swarm-status-command-plan.md @@ -0,0 +1,101 @@ +# Implementation Plan: /swarm-status Command + +**Date:** 2026-01-25 +**Feature:** Add a `/swarm-status` command that shows the current status of any active swarm teams + +## Overview + +Create a simple read-and-display command that shows the status of active swarm teams by reading team configuration files from `~/.claude/teams/` and task files from `~/.claude/tasks/`. + +## Implementation + +### File to Create + +`plugins/compound-engineering/commands/swarm-status.md` + +### Command Structure + +```yaml +--- +name: swarm-status +description: Show the current status of active swarm teams, including members and task progress +argument-hint: "[optional: team-name]" +--- +``` + +### Command Behavior + +1. **Discover Active Teams** + - Read all directories in `~/.claude/teams/` + - For each team, read `config.json` to get team metadata + +2. **Display Team Information** + - Team name and description + - Creation date + - Team lead + - List of members with their role/color + +3. **Display Task Progress** + - Read task files from `~/.claude/tasks/{team-name}/` + - Show task summary: completed/total + - List tasks with their status and owner + +4. **Optional: Filter by Team Name** + - If `$ARGUMENTS` provided, show only that specific team + +### Output Format + +```markdown +## Active Swarm Teams + +### Team: slfg-20260125-193700 +**Description:** Swarm for: Add /swarm-status command +**Created:** 2026-01-25 19:37:00 +**Lead:** team-lead + +#### Members (3) +| Name | Role | Color | +|------|------|-------| +| team-lead | team-lead | - | +| planner | - | blue | +| researcher | - | green | + +#### Tasks (2/8 complete) +| ID | Subject | Status | Owner | +|----|---------|--------|-------| +| 1 | Create implementation plan | in_progress | planner | +| 2 | Research best practices | in_progress | researcher | +| 3 | Implement the approved plan | pending | - | +... + +--- + +**Total Teams:** 1 | **Total Active Agents:** 3 +``` + +### Command Implementation Steps + +1. Use bash to list `~/.claude/teams/*/config.json` +2. Parse each config.json with jq or read tool +3. For each team, read task files from `~/.claude/tasks/{team-name}/` +4. Format output as markdown table +5. If no teams found, display "No active swarm teams" + +### Error Handling + +- If `~/.claude/teams/` doesn't exist: "No active swarm teams" +- If specific team not found: "Team '{name}' not found" +- If task directory missing: Show team info without tasks + +## Checklist + +- [ ] Create `plugins/compound-engineering/commands/swarm-status.md` +- [ ] Update plugin.json description with new command count +- [ ] Update marketplace.json description with new command count +- [ ] Update README.md command list +- [ ] Update CHANGELOG.md +- [ ] Run `/release-docs` to update documentation site + +## Estimated Complexity + +**Low** - This is a straightforward read-and-display command with no complex logic. Similar to how other status-reporting commands work in the codebase.