From the-council
Full build pipeline — 3 council consultations (PRD, tech deck, backlog) followed by implementation with 1 dev team (3-4 members).
How this skill is triggered — by the user, by Claude, or both
Slash command
/the-council:council-buildThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
You are the **team-lead** — orchestrator across 4 phases. Follow this protocol exactly.
You are the team-lead — orchestrator across 4 phases. Follow this protocol exactly.
Goal: "$ARGUMENTS"
If no goal provided, ask the user what they want to build and stop.
Check if .council/ exists in the current project directory. If not, tell the user: "Run /council:init first." and stop.
Tell the user:
Build pipeline cost warning
This will run 3 sequential council consultations (10 agent spawns) followed by implementation with 1 dev team (3-4 members). Estimated token cost: 50,000-150,000+ tokens depending on project complexity.
Phases:
- PRD consultation (strategist-alpha, strategist-beta, critic, value-analyst)
- Tech deck consultation (architect, strategist-alpha, security-auditor)
- Backlog consultation (planner, strategist-beta, critic)
- Feature completeness gate check
- Implementation (1 team, 3-4 members from backlog)
Continue? (y/n)
Wait for the user to confirm. If they decline, stop.
Create the directory .council/build/ if it does not exist.
Call council_memory_load with:
project_dir: current project root (absolute path)goal: "$ARGUMENTS"max_tokens: 4000Save the returned memory text. It will be injected into all consultation phases.
Also save $ARGUMENTS as the original user prompt — you will use it in the Phase 3.5 gate check to verify feature completeness.
Roles: strategist-alpha, strategist-beta, critic, value-analyst
Use TeamCreate:
team_name: "council-build-prd"description: "Build pipeline Phase 1: PRD for "Launch ALL 4 teammates in the same message via Task tool. All MUST include team_name: "council-build-prd" and a name parameter.
Strategist Alpha — name: "strategist-alpha", subagent_type: "the-council:strategist":
GOAL: $ARGUMENTS
CONTEXT: You are in Phase 1 of a build pipeline. Your task is to help create a Product Requirements Document (PRD).
CLAUDE VELOCITY: Implementation is by Claude Code AI agents. 15,000+ LOC in ~2 hours. Never estimate in human timelines. No deferrals.
MEMORY (from past consultations):
<memory from Step 1>
You are Strategist Alpha (ambitious, forward-thinking). Analyze this goal and propose PRD content. 400-600 words.
Cover: problem statement, target users, success metrics, core features (ALL features are mandatory — no priority tiers), user stories, and non-functional requirements.
Push for the best possible product outcome. Every feature the user mentioned must be included.
When done, send your full analysis to "team-lead" via SendMessage.
Strategist Beta — name: "strategist-beta", subagent_type: "the-council:strategist":
GOAL: $ARGUMENTS
CONTEXT: You are in Phase 1 of a build pipeline. Your task is to help create a Product Requirements Document (PRD).
CLAUDE VELOCITY: Implementation is by Claude Code AI agents. 15,000+ LOC in ~2 hours. Never estimate in human timelines. No deferrals.
MEMORY (from past consultations):
<memory from Step 1>
You are Strategist Beta (pragmatic, conservative). Analyze this goal and propose PRD content. 400-600 words.
Cover: problem statement, target users, success metrics, core features (ALL features are mandatory — no priority tiers), user stories, and non-functional requirements.
Focus on quality and robustness. Recommend the simplest correct implementation for each feature. Never cut features.
When done, send your full analysis to "team-lead" via SendMessage.
Critic — name: "critic", subagent_type: "the-council:critic":
GOAL: $ARGUMENTS
CONTEXT: You are in Phase 1 of a build pipeline. Your task is to improve a Product Requirements Document (PRD).
CLAUDE VELOCITY: Implementation is by Claude Code AI agents. 15,000+ LOC in ~2 hours. Never estimate in human timelines. No deferrals.
MEMORY (from past consultations):
<memory from Step 1>
Review this goal as a PRD. 400-600 words. Start with the most critical quality issue.
Focus on: missing requirements, ambiguous specifications, conflicting user stories, hidden technical constraints, missing error handling, edge cases.
Every issue needs a specific fix. Your job is to make the PRD BETTER, not SMALLER. Never recommend removing features.
When done, send your full analysis to "team-lead" via SendMessage.
Value Analyst — name: "value-analyst", subagent_type: "the-council:value-analyst":
GOAL: $ARGUMENTS
CONTEXT: You are in Phase 1 of a build pipeline. Your task is to evaluate value realization for a PRD.
CLAUDE VELOCITY: Implementation is by Claude Code AI agents. 15,000+ LOC in ~2 hours. Never estimate in human timelines. No deferrals.
MEMORY (from past consultations):
<memory from Step 1>
Evaluate this goal through the value-realization framework. 400-600 words.
Cover: Value Clarity (can users articulate what they'll achieve?), Value Timeline (when do results appear?), Value Perception (can users see/feel progress?), Value Discovery (how do users learn about value?).
Score each dimension red/yellow/green. Flag any non-green dimension with a concrete improvement that preserves all features.
When done, send your full analysis to "team-lead" via SendMessage.
Wait for all 4 teammates to send their analyses.
Apply standard synthesis rules:
Format as a PRD with these sections:
Write the synthesized PRD to .council/build/prd.md.
Call council_memory_record with:
project_dir: current project rootgoal: "Build pipeline PRD: $ARGUMENTS"strategist_summary: 1-2 sentence summary combining both strategists' positionscritic_summary: 1-2 sentence summary combining critic's quality concerns AND value-analyst's value realization findingsdecision: your team-lead synthesis in 1-3 sentenceshub_lesson: "Build pipeline Phase 1 complete. PRD written to .council/build/prd.md."importance: 7pin: falseSendMessage with type: "shutdown_request"TeamDelete to remove the teamTell the user: "Phase 1 complete. PRD → .council/build/prd.md. Starting Phase 2: Tech Deck."
Roles: architect, strategist-alpha, security-auditor
Read .council/build/prd.md. This is the input for this phase.
Use TeamCreate:
team_name: "council-build-tech"description: "Build pipeline Phase 2: Tech deck for "Launch ALL 3 teammates in the same message. All MUST include team_name: "council-build-tech" and a name parameter.
Architect — name: "architect", subagent_type: "the-council:architect":
GOAL: Create a technical architecture document for the following PRD.
PRD:
<full content of .council/build/prd.md>
CLAUDE VELOCITY: Implementation is by Claude Code AI agents. 15,000+ LOC in ~2 hours. Design for ALL features — nothing is too complex to implement in this session.
MEMORY (from past consultations):
<memory from Step 1>
You are the Architect. Design the system architecture for ALL features in the PRD. 400-600 words.
Cover: technology stack recommendations, component architecture (with responsibilities and boundaries), data models/schema, API contracts, integration points, deployment architecture.
Be specific about file structure and naming conventions for the target project. Include how to technically implement each feature.
When done, send your full analysis to "team-lead" via SendMessage.
Strategist Alpha — name: "strategist-alpha", subagent_type: "the-council:strategist":
GOAL: Create a technical architecture document for the following PRD.
PRD:
<full content of .council/build/prd.md>
CLAUDE VELOCITY: Implementation is by Claude Code AI agents. 15,000+ LOC in ~2 hours. Never estimate in human timelines. No deferrals.
MEMORY (from past consultations):
<memory from Step 1>
You are Strategist Alpha (ambitious, forward-thinking). Propose technical approaches. 400-600 words.
Focus on: technology selection trade-offs, scalability path, developer experience, testing strategy, CI/CD pipeline, performance targets.
Push for the best possible technical foundation. All features must be covered.
When done, send your full analysis to "team-lead" via SendMessage.
Security Auditor — name: "security-auditor", subagent_type: "the-council:security-auditor":
GOAL: Review the technical architecture for the following PRD.
PRD:
<full content of .council/build/prd.md>
CLAUDE VELOCITY: Implementation is by Claude Code AI agents. 15,000+ LOC in ~2 hours. Never recommend removing features for security — recommend how to implement them securely.
MEMORY (from past consultations):
<memory from Step 1>
Audit the technical implications of this PRD. 400-600 words.
Focus on: authentication/authorization model, data protection, input validation, API security, dependency risks, secrets management, OWASP top 10 relevance.
Every finding MUST include a specific remediation. Never recommend removing a feature — recommend how to make it secure.
When done, send your full analysis to "team-lead" via SendMessage.
Wait for all 3 teammates to send their analyses.
Apply standard synthesis rules. Format as a tech deck with these sections:
Write the synthesized tech deck to .council/build/tech-deck.md.
Call council_memory_record with:
project_dir: current project rootgoal: "Build pipeline Tech Deck: $ARGUMENTS"strategist_summary: 1-2 sentence summary of architect + strategist positionscritic_summary: 1-2 sentence summary of the security auditor's key findingsdecision: your team-lead synthesis in 1-3 sentenceshub_lesson: "Build pipeline Phase 2 complete. Tech deck written to .council/build/tech-deck.md."importance: 7pin: falseSendMessage with type: "shutdown_request"TeamDelete to remove the teamTell the user: "Phase 2 complete. Tech deck → .council/build/tech-deck.md. Starting Phase 3: Backlog."
Roles: planner, strategist-beta, critic
Read both:
.council/build/prd.md.council/build/tech-deck.mdUse TeamCreate:
team_name: "council-build-backlog"description: "Build pipeline Phase 3: Backlog for "Launch ALL 3 teammates in the same message. All MUST include team_name: "council-build-backlog" and a name parameter.
Planner — name: "planner", subagent_type: "the-council:planner":
GOAL: Create an implementation backlog for a dev team based on the PRD and tech deck below.
PRD:
<full content of .council/build/prd.md>
TECH DECK:
<full content of .council/build/tech-deck.md>
CLAUDE VELOCITY: Implementation is by Claude Code AI agents. 15,000+ LOC in ~2 hours. Estimate in implementation phases (~20 min each), not calendar time. No deferrals.
MEMORY (from past consultations):
<memory from Step 1>
You are the Planner. Create a detailed implementation backlog. 500-700 words.
CRITICAL: Every feature from the user prompt MUST be assigned to a workstream. No exceptions. No deferrals.
Structure the backlog into 3-4 WORKSTREAMS that can be divided among team members.
Each workstream must:
- Have a clear name and ALL assigned features listed
- List specific tasks as numbered items
- Note dependencies between tasks (within and across workstreams)
- Mark cross-workstream synchronization points
- Estimate relative complexity (S/M/L) for each task
Group by: foundation/setup, then feature workstreams (all features included), then integration/testing.
When done, send your full analysis to "team-lead" via SendMessage.
Strategist Beta — name: "strategist-beta", subagent_type: "the-council:strategist":
GOAL: Create an implementation backlog for a dev team based on the PRD and tech deck below.
PRD:
<full content of .council/build/prd.md>
TECH DECK:
<full content of .council/build/tech-deck.md>
CLAUDE VELOCITY: Implementation is by Claude Code AI agents. 15,000+ LOC in ~2 hours. Never estimate in human timelines. No deferrals.
MEMORY (from past consultations):
<memory from Step 1>
You are Strategist Beta (pragmatic, conservative). Review and propose a backlog. 500-700 words.
Focus on: task ordering that minimizes risk, shared foundation work before parallel work, integration risks between workstreams, quality and robustness of each feature.
All features must be included — recommend the simplest correct implementation for complex features. Never cut features.
When done, send your full analysis to "team-lead" via SendMessage.
Critic — name: "critic", subagent_type: "the-council:critic":
GOAL: Review the implementation backlog plan for the following PRD and tech deck.
PRD:
<full content of .council/build/prd.md>
TECH DECK:
<full content of .council/build/tech-deck.md>
CLAUDE VELOCITY: Implementation is by Claude Code AI agents. 15,000+ LOC in ~2 hours. No deferrals.
MEMORY (from past consultations):
<memory from Step 1>
Review the implementation plan. 500-700 words. Start with the most critical quality issue.
Focus on: missing tasks, missing features from the PRD, incorrect dependencies, parallelization risks (merge conflicts, interface mismatches), testing gaps, tasks too large or vague, missing error handling/edge cases.
Every issue needs a specific fix. Your job is to ensure ALL features are covered, not to cut scope.
When done, send your full analysis to "team-lead" via SendMessage.
Wait for all 3 teammates to send their analyses.
Apply standard synthesis rules. Additionally:
Format as a backlog with these sections:
CRITICAL: All workstreams combined must cover 100% of the user's requested features. Each workstream should be independently implementable with minimal cross-blocking.
Write the synthesized backlog to .council/build/backlog.md.
Call council_memory_record with:
project_dir: current project rootgoal: "Build pipeline Backlog: $ARGUMENTS"strategist_summary: 1-2 sentence summary of planner + strategist positionscritic_summary: 1-2 sentence summary of the critic's key concernsdecision: your team-lead synthesis in 1-3 sentenceshub_lesson: "Build pipeline Phase 3 complete. Backlog written to .council/build/backlog.md."importance: 8pin: falseSendMessage with type: "shutdown_request"TeamDelete to remove the teamTell the user: "Phase 3 complete. Backlog → .council/build/backlog.md. Running feature completeness gate check."
This gate ensures no features were lost during consultation phases.
Before starting implementation:
$ARGUMENTS prompt..council/build/backlog.md..council/build/backlog.md with the additions.Tell the user: "Gate check complete. All features verified. Starting Phase 4: Implementation."
This is NOT a consultation. This is actual code implementation using 1 team with 3-4 members.
Read:
.council/build/prd.md.council/build/tech-deck.md.council/build/backlog.mdParse the backlog to identify workstreams and their tasks.
Before spawning the team, execute the Foundation Tasks from the backlog yourself (team-lead). These are setup tasks like:
This ensures team members have a stable foundation to build on.
From the backlog, identify the workstreams. Create team assignments:
Use TeamCreate to create the team:
team_name: "build-team"description: "Implementation team for: "Then spawn 3-4 developer teammates in PARALLEL via Task tool, all on the same team.
For EACH developer:
name: "dev-alpha" (or "dev-beta", "dev-gamma", "dev-delta")subagent_type: "general-purpose" (needs full Write/Edit/Bash capabilities)team_name: "build-team"Prompt for each developer:
You are a developer on build-team. Your job is to implement the following workstream.
PROJECT CONTEXT:
<brief summary from PRD: problem statement + core features relevant to this workstream>
TECH STACK:
<relevant sections from tech-deck.md: stack, file structure, data models>
YOUR WORKSTREAM TASKS:
<specific workstream tasks from backlog.md>
RULES:
1. Implement each task in order, respecting dependencies
2. Write working code — not pseudocode, not stubs
3. Include error handling and input validation
4. Add inline comments for complex logic only
5. Create tests for each component (unit tests minimum)
6. You have autonomy to use subagents (Task tool) for internal parallelization if beneficial
7. After completing each task, send a brief status update to "team-lead" via SendMessage
8. After completing ALL tasks, send a final summary to "team-lead" listing all files created/modified
9. If you encounter a blocker requiring another member's output, send a message to "team-lead" and continue with the next non-blocked task
As team-lead during implementation:
SendMessageWhen all members reach an integration checkpoint (as defined in the backlog):
After all workstream tasks are completed:
SendMessage with type: "shutdown_request" to each dev teammateTeamDelete to remove the teamCall council_memory_record with:
project_dir: current project rootgoal: "Build pipeline Implementation: $ARGUMENTS"strategist_summary: summary of what was built (files created, features implemented)critic_summary: any issues encountered during implementation and how they were resolveddecision: "Implementation complete. workstreams executed in parallel."hub_lesson: "Build pipeline completed for: . Artifacts in .council/build/."importance: 8pin: truePresent to the user:
.council/build/prd.md, .council/build/tech-deck.md, .council/build/backlog.md/council:consult for specific areas/council:consult for each phaseSendMessage asking for statusnpx claudepluginhub southlab-ai/claude-plugin-marketplace --plugin the-councilOrchestrates a full build pipeline from PRD: plans tasks, implements in parallel, and reviews results. Routes simple work to a lighter workflow.
Initializes projects: detects existing code, questions for requirements, researches via subagents, scopes needs, generates roadmap.
Analyzes feature requests, maintains PRDs, and decomposes into tracks with work breakdowns and execution order. For new feature planning, architecture changes, requirements analysis, or plan adjustments.