Design exploration using parallel agents through a 7-phase process: topic analysis, memory context, divergent ideation (10+ ideas), feasibility filtering, evaluation with devil's advocate scoring (0-10 across 7 dimensions), synthesis of top approaches, and trade-off comparison. Supports open exploration, constrained design, comparison, quick ideation, and iterative optimization modes. Use when brainstorming ideas, exploring solutions, or comparing alternatives.
Design exploration using parallel agents through a 7-phase process: topic analysis, memory context, divergent ideation (10+ ideas), feasibility filtering, evaluation with devil's advocate scoring (0-10 across 7 dimensions), synthesis of top approaches, and trade-off comparison. Supports open exploration, constrained design, comparison, quick ideation, and iterative optimization modes. Use when brainstorming ideas, exploring solutions, or comparing alternatives.
{"keywords":["brainstorm","brianstorm","explore ideas","explore some ideas","ideate","think through","some approaches","not sure how to architect","alternatives","compare strategies","help me compare"],"examples":["brainstorm how to build the notification system","what are some approaches for handling real-time updates","im not sure how to architect the billing module"],"anti-triggers":["implement","build","fix","commit","review","assess","rate"]}
Brainstorming Ideas Into Designs
Transform rough ideas into fully-formed designs through intelligent agent selection and structured exploration.
Core principle: Analyze the topic, select relevant agents dynamically, explore alternatives in parallel, present design incrementally.
Argument Resolution
TOPIC = "$ARGUMENTS"# Full argument string, e.g., "API design for payments"# $ARGUMENTS[0] is the first token (CC 2.1.59 indexed access)
STEP -1: MCP Probe + Resume Check
Probe MCP servers once at skill start, store capabilities, and resume from any prior crashed session. Each phase emits a JSON handoff file consumed by the next.
Full procedure + handoff-file table: Read("${CLAUDE_PLUGIN_ROOT}/skills/brainstorm/references/mcp-probe-resume.md")
STEP 0: Project Context Discovery
BEFORE creating tasks or selecting agents, detect the project tier. This becomes the complexity ceiling for all downstream decisions.
Auto-Detection (scan codebase)
# PARALLEL — quick signals (launch all in ONE message)
Grep(pattern="take-home|assignment|interview|hackathon", glob="README*", output_mode="content")
Grep(pattern="take-home|assignment|interview|hackathon", glob="*.md", output_mode="content")
Glob(pattern=".github/workflows/*")
Glob(pattern="**/Dockerfile")
Glob(pattern="**/terraform/**")
Glob(pattern="**/k8s/**")
Glob(pattern="CONTRIBUTING.md")
Pass the detected tier as context to ALL downstream agents and phases. The tier constrains which patterns are appropriate — see scope-appropriate-architecture skill for the full matrix.
Override: User can always override the detected tier. Warn them of trade-offs if they choose a higher tier than detected.
STEP 0a: Verify User Intent with AskUserQuestion
Clarify brainstorming constraints:
# NOTE: AskUserQuestion caps each question at 4 options (CC schema: minItems 2,# maxItems 4). Use plain `label` + `description` only — the schema permits a# `preview` field, but on current CC it forces a side-by-side picker layout with# dead up/down keyboard nav (confirmed 2026-05-28), so skills no longer use it# (`markdown` was never valid). The 6 legacy# modes are split across 3 valid questions: Q1 = exploration flow, Q2 folds the# old "Constrained design" mode into constraints, Q3 carries the orthogonal# "Plan first" preamble (it composes with any Q1 mode — it was never mutually# exclusive). "Quick ideation" + STEP 0c /effort=low overlap; both downscale.
AskUserQuestion(
questions=[
{
"question": "What type of design exploration?",
"header": "Mode",
"options": [
{"label": "Open exploration (Recommended)", "description": "Generate 10+ ideas, evaluate all, synthesize top 3"},
{"label": "Comparison", "description": "Compare 2-3 specific approaches I have in mind"},
{"label": "Quick ideation", "description": "Generate ideas fast, skip deep evaluation"},
{"label": "Iterative optimization", "description": "Try, measure, keep/discard, repeat (autoresearch-style)"}
],
"multiSelect": false
},
{
"question": "Any preferences or constraints?",
"header": "Constraints",
"options": [
{"label": "None", "description": "Explore all possibilities"},
{"label": "Use existing patterns", "description": "Prefer patterns already in codebase"},
{"label": "Minimize complexity", "description": "Favor simpler solutions"},
{"label": "Fixed requirements (constrained)", "description": "Hard requirements to work within — skip divergent phase, focus on feasibility (old 'Constrained design' mode)"}
],
"multiSelect": false
},
{
"question": "Research before ideating?",
"header": "Plan-first",
"options": [
{"label": "No — dive straight in (Recommended)", "description": "Go directly to ideation"},
{"label": "Yes — plan first", "description": "Read-only research (EnterPlanMode): scan codebase, map the solution space, then ExitPlanMode for approval before Phase 1 (old 'Plan first' mode)"}
],
"multiSelect": false
}
]
)
If Q3 = 'Yes — plan first' (composes with any Q1 mode):
# 1. Enter read-only plan mode
EnterPlanMode("Brainstorm exploration: $TOPIC")
# 2. Research phase — Read/Grep/Glob ONLY, no Write/Edit# - Scan existing codebase for related patterns# - Search for prior decisions on this topic (memory graph)# - Identify constraints, dependencies, and trade-offs# 3. Produce structured exploration plan:# - Key questions to answer# - Dimensions to explore# - Agents to spawn and their focus areas# - Evaluation criteria# 4. Exit plan mode — returns plan for user approval
ExitPlanMode()
# 5. User reviews. If approved → continue to Phase 1 with plan as input.
Based on answers, adjust workflow:
Open exploration (Q1): Full 7-phase process with all agents
Comparison (Q1): Skip ideation, jump to evaluation phase
Quick ideation (Q1): Generate ideas, skip deep evaluation
Iterative optimization (Q1): Skip phases 2-6, enter autoresearch-style loop (see below)
Constrained design (Q2 = "Fixed requirements"): Skip divergent phase, focus on feasibility within the stated requirements — composes with any Q1 mode
If 'Iterative optimization' selected: skip Phases 2-6 and enter the autoresearch-style metric-driven loop.
Full sub-flow (metric question, baseline, loop body): Read("${CLAUDE_PLUGIN_ROOT}/skills/brainstorm/references/iterative-optimization-mode.md")
STEP 0b: Select Orchestration Mode (skip for Tier 1-2)
Choose Agent Teams (mesh — agents debate and challenge ideas) or Task tool (star — all report to lead):
Agent Teams mode (GA since CC 2.1.33) → recommended for 3+ agents (real-time debate produces better ideas)
Read the /effort setting and scale brainstorm depth — low runs phases 0/2/5 only, high (default) runs all 7, xhigh adds extra devil's-advocate and synthesis rounds. Explicit user choice in STEP 0a always overrides downscaling.
Full level table + detection rules: Read("${CLAUDE_PLUGIN_ROOT}/skills/brainstorm/references/effort-scaling.md")
🚨 CRITICAL: Task Management is MANDATORY (CC 2.1.16)
# 1. Create main task IMMEDIATELY
TaskCreate(
subject="Brainstorm: {topic}",
description="Design exploration with parallel agent research",
activeForm="Brainstorming {topic}"
)
# 2. Create subtasks for each phase
TaskCreate(subject="Analyze topic and select agents", activeForm="Analyzing topic") # id=2
TaskCreate(subject="Search memory for past decisions", activeForm="Searching knowledge graph") # id=3
TaskCreate(subject="Generate divergent ideas (10+)", activeForm="Generating ideas") # id=4
TaskCreate(subject="Feasibility fast-check", activeForm="Checking feasibility") # id=5
TaskCreate(subject="Evaluate with devil's advocate", activeForm="Evaluating ideas") # id=6
TaskCreate(subject="Synthesize top approaches", activeForm="Synthesizing approaches") # id=7
TaskCreate(subject="Present design options", activeForm="Presenting options") # id=8# 3. Set dependencies (sequential chain: 2→3→4→5→6→7→8)for i inrange(3, 9):
TaskUpdate(taskId=str(i), addBlockedBy=[str(i-1)])
# 4. Update status as you progress
TaskUpdate(taskId="2", status="in_progress") # When starting
TaskUpdate(taskId="2", status="completed") # When done — repeat for each subtask
Filter to top 2-3, trade-off table, test strategy per approach
Options
6. Design Presentation
Present in 200-300 word sections, log to experiment journal
Validated design
6.4. Living Plan Playground (optional, on request or multi-wave plan)
Emit an updatable LPP playground via visualize-plan's living-plan path
<slug>.html (living)
6.5. Audio Podcast (signal-fired, optional)
Auto-emit brainstorm-podcast.m4a when composite≥8.0 + approaches≥4
.m4a file
Phase 6.4 — Living Plan Playground (optional)
When the user asked for a playground output, OR the synthesis produced a multi-wave plan (2+
execution waves with verifiable items), emit the design as a living plan playground instead of a
one-shot render: waves from Phase 5 synthesis, 7-dimension scores from Phase 4, discarded ideas with
reasons from Phase 3, one item per work unit with a "done when" evidence check. Route through
visualize-plan's living-plan path (exemplar living-plan.template.html, update-mode contract in its
references/format-dispatch.md §Living-plan update mode). The artifact then tracks its own execution:
later sessions flip item statuses in the embedded lpp-state JSON and append to its changelog — one
plan = one file, never fork.
After Phase 6 lands, optionally invoke scripts/post_synth_podcast.py <session-dir> to auto-emit an audio podcast summarizing the top approaches + trade-offs. Self-skips on every non-happy-path so it never breaks the brainstorm:
Auto-skip conditions (all exit 0, all WARN-logged):
Skip reason
Trigger
signal absent
composite[recommended] < 8.0 OR len(approaches) < 4
design-doc.md not found
Phase 6 didn't write a design doc to the session dir
yg-mcp-core not importable
yg-mcp-core>=0.3.0 not installed (orchestkit is public; yg-mcp-core lives on private pypi.yonyon.ai — HQ-only)
hq-content MCP unreachable
MCP server down OR .mcp.json doesn't define hq-content
Session dir must contain brainstorm-output.json (with recommended, composite, approaches) + design-doc.md. Handoff JSON at <session-dir>/brainstorm-podcast.json records status (fired / skipped) and m4a_path on success.
Mirrors Yonatan-HQ/hq-ext-plugin#194 (audio_podcast handler for /hq-ext:decide).
Progressive Output (CC 2.1.76)
Output results incrementally after each phase — don't batch everything until the end:
Each agent's ideas as they return (don't wait for all)
3. Keep/Discard Gate
Survivors and discard reasons (keep/discard/crash per idea)
4. Evaluation
Top-rated ideas with 7-dimension scores
For Phase 2 parallel agents, output each agent's ideas as soon as it returns — don't wait for all agents. This lets users see early ideas and redirect the exploration if needed. Showing ideas incrementally also helps users build a mental model of the solution space faster than a final dump.
Load the phase workflow for detailed instructions:
Always include:workflow-architect for system design perspective, test-generator for testability assessment.
Agent Teams Alternative: Brainstorming Team
In Agent Teams mode, form a brainstorming team where agents debate ideas in real-time. Dynamically select teammates based on topic analysis (Phase 0):
# CC 2.1.178+: one implicit team per session — no TeamCreate.# Spawn teammates directly via Agent(name=...). Requires# CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 (set in ork.settings.json).# Always include the system design lead
Agent(subagent_type="ork:workflow-architect", name="system-designer",
team_name="brainstorm-{topic-slug}",
prompt="""You are the system design lead for brainstorming: {topic}
DIVERGENT MODE: Generate 3-4 architectural approaches.
When other teammates share ideas, build on them or propose alternatives.
Challenge ideas that seem over-engineered — advocate for simplicity.
After divergent phase, help synthesize the top approaches.""")
# Domain-specific teammates (select 2-3 based on topic keywords)
Agent(subagent_type="ork:backend-system-architect", name="backend-thinker",
team_name="brainstorm-{topic-slug}",
prompt="""Brainstorm backend approaches for: {topic}
DIVERGENT MODE: Generate 3-4 backend-specific ideas.
When system-designer shares architectural ideas, propose concrete API designs.
Challenge ideas from other teammates with implementation reality checks.
Play devil's advocate on complexity vs simplicity trade-offs.""")
Agent(subagent_type="ork:frontend-ui-developer", name="frontend-thinker",
team_name="brainstorm-{topic-slug}",
prompt="""Brainstorm frontend approaches for: {topic}
DIVERGENT MODE: Generate 3-4 UI/UX ideas.
When backend-thinker proposes APIs, suggest frontend patterns that match.
Challenge backend proposals that create poor user experiences.
Advocate for progressive disclosure and accessibility.""")
# Always include: testability assessor
Agent(subagent_type="ork:test-generator", name="testability-assessor",
team_name="brainstorm-{topic-slug}",
prompt="""Assess testability for each brainstormed approach: {topic}
For every idea shared by teammates, evaluate:
- Can core logic be unit tested without external services?
- What's the mock/stub surface area?
- Can it be integration-tested with docker-compose/testcontainers?
Score testability 0-10 per the evaluation rubric.
Challenge designs that score below 5 on testability.
Propose test strategies for the top approaches in synthesis phase.""")
# Optional: Add security-auditor, llm-integrator based on topic
Key advantage: Agents riff on each other's ideas and play devil's advocate in real-time, rather than generating ideas in isolation.
Fork pattern (CC 2.1.89 — #1227): All brainstorm agents are fork-eligible: prompts are <500 words, no custom model, no worktree. CC shares the parent's cached API prefix across forks, reducing cost by ~60%. Do NOT add model= to agent calls. See chain-patterns/references/fork-pattern.md.
Nested delegation (CC 2.1.172+): Teammates MAY be instructed to delegate a bounded sub-question to their declared sub-agents (e.g. backend-thinker → database-engineer to sanity-check a schema idea) instead of guessing inline. Keep chains ≤ 3 levels deep; divergent ideation itself stays flat — the parallel teammates above ARE the fan-out for independent ideas. See chain-patterns Pattern 9 (CC 2.1.172+).
Team teardown after synthesis:
# After Phase 5 synthesis and design presentation# CC 2.1.178+: no TeamDelete — teammates wind down at turn end# (press Ctrl+F twice to stop lingering background teammates).# Worktree cleanup (CC 2.1.72) — for Tier 3+ projects that entered a worktree# If EnterWorktree was called during brainstorm (e.g., Plan first → worktree), exit it
ExitWorktree(action="keep") # Keep branch for follow-up /ork:implement
Fallback: If team formation fails, load Read("${CLAUDE_PLUGIN_ROOT}/skills/brainstorm/references/phase-workflow.md") and use standard Phase 2 Task spawns.
Partial results (CC 2.1.76): Background agents that are killed (timeout, context limit) return responses tagged with [PARTIAL RESULT]. When collecting Phase 2 divergent ideas, check each agent's output for this tag. If present, include the partial ideas but note them as incomplete in Phase 3 feasibility. Prefer synthesizing partial results over re-spawning agents.
PostCompact recovery: Long brainstorm sessions may trigger context compaction. The PostCompact hook re-injects branch and task state. If compaction occurs mid-brainstorm, check .claude/chain/state.json for the last completed phase and resume from the next handoff file (see Phase Handoffs table). Reactive compaction (CC 2.1.142+) now sizes the first summarize to the actual overflow, so mid-turn stalls are rare — no need to expect a second pass.
When context fills (CC 2.1.141+): Instead of ending the session, use the rewind menu's "Summarize up to here" to compress earlier turns while keeping recent context intact — a deliberate compaction point that pairs with the improved reactive compaction above.
Session recap (CC 2.1.108+): After idle periods, use /recap to restore conversational context alongside checkpoint-resume. Enabled by default since CC 2.1.110 (even with telemetry disabled).
Push notifications (CC 2.1.110+): For long brainstorm sessions, use PushNotification to alert when synthesis is complete (requires Remote Control with "Push when Claude decides").
Lighter alternative (CC 2.1.141+): Hooks can emit desktop notifications, window titles, and bells natively via the terminalSequence field in hook JSON output — no Remote Control required, no Anthropic round-trip. Ork's notification hooks (notification/desktop, notification/sound) already emit via terminalSequence (#1847, shipped).
Manual cleanup: To stop lingering background teammates, press Ctrl+F twice (foreground/Stop). Note: /clear (CC 2.1.72+) preserves background agents — only foreground tasks are cleared.
Teammate background tasks survive turn-end (CC 2.1.183): A background task started by a teammate is no longer killed when that teammate finishes its turn. Phase-2 teammates may safely kick off a long run_in_background task (e.g. a feasibility build) and let it outlive their own turn — the result is still collectable at synthesis. Pre-2.1.183 this required the lead to own the background task.
Key Principles
Principle
Application
Dynamic agent selection
Select agents based on topic keywords
Parallel research
Launch 3-5 agents in ONE message
Memory-first
Check graph for past decisions before research
Divergent-first
Generate 10+ ideas BEFORE filtering
Task tracking
Use TaskCreate/TaskUpdate for progress visibility
YAGNI ruthlessly
Remove unnecessary complexity
Running unattended with /goal
Set a completion condition with /goal (CC 2.1.139+, 2.1.143+ recommended for this skill) and this skill will keep working across turns until the condition is met. Works in interactive, -p, and Remote Control. The overlay panel shows live elapsed / turns / tokens.
Why 2.1.143+: Pre-2.1.143 /goal evaluator could fire while background shells or delegated subagents were still running — racing the parallel agents this skill spawns in Phase 2. Fixed in CC 2.1.143 (changelog: "/goal evaluator firing while background shells or delegated subagents are still running").
Example completion condition for this skill:
/goal until design.options_count >= 2 AND user_chose_option, or stop after 8 turns
Stops when: 2+ ranked design options presented and the user selects one (or after Phase 6 if running unattended via -p). Compatible with claude.ai Remote Control runs.
Quality Bar
Done means all of these hold:
Phase 2 produced >=10 distinct ideas BEFORE any filtering, each surfaced as its agent returned
Every idea past the keep/discard gate carries a one-line keep/discard/crash verdict
Each surviving idea scores 0-10 on all 7 dimensions (simplicity included) plus one devil's-advocate challenge
Synthesis narrows to 2-3 approaches with a trade-off table AND a per-approach test strategy
Output ends with ranked options the user chooses from (or Phase 6 completes when running unattended via -p)
📜 Related Skills
ork:architecture-decision-record - Document key decisions made during brainstorming
ork:implement - Execute the implementation plan after brainstorming completes
ork:explore - Deep codebase exploration to understand existing patterns
ork:assess - Rate quality 0-10 with dimension breakdown
ork:design-to-code - Convert brainstormed UI designs into components
ork:component-search - Find existing components before building new ones
ork:competitive-analysis - Porter's Five Forces, SWOT for product brainstorms
Picker fallback (#1795)
The picker stall reported in orchestkit#1795 was a schema break, not a CC input bug: questions with >4 options or a markdown field (the markdown key is not valid — the schema field is preview) fail AskUserQuestion validation, so the picker never renders. All skill questions conform to the schema (2–4 options, no preview on multiSelect), enforced by tests/skills/structure/test-askuserquestion-schema.sh. Separately, although the schema permits a preview field, current CC renders any question that uses one in a side-by-side picker layout where ↑/↓ keyboard navigation is dead (confirmed 2026-05-28). So skills now standardize on plain label + description only and no longer set preview. If you still hit a stall on a future CC build, ORK_ASK_FALLBACK=text remains as a defensive opt-in: the lifecycle/ask-fallback-injector hook then tells the assistant to pose options inline as a numbered list. Hook propagates globally — no per-skill edit needed.
References
Load on demand with Read("${CLAUDE_PLUGIN_ROOT}/skills/brainstorm/references/<file>"):
