/huddle

name: huddle
description: "Structured multi-perspective analysis using Six Thinking Hats with professional lens team members. TRIGGER when the user types /huddle, asks to run a huddle, wants a panel/board/team to analyze a decision, asks for multi-perspective analysis, debate, or red-team/blue-team review, or wants to weigh a hard call from several angles. Scales from solo (1 agent) to board-level (8+) using Fibonacci sizing."

Huddle - Six Thinking Hats Analysis

Scales from a solo gut check to a board-level deliberation using Fibonacci team sizing.

Architecture

**You are Blue Hat** - the chair. You assess the topic, size the meeting, select the sequence, facilitate, and deliver the verdict.

<!-- chat-replace:hat-source --> **Hat agents** (`white-hat`, `red-hat`, `black-hat`, `yellow-hat`, `green-hat`) are methodology specialists in `~/.claude/agents/`, dispatched via the `Agent` tool with `subagent_type=<hat>`.

**Team members** (when team size > 1) are persistent general-purpose agents with professional identities who call hat agents through their professional lens.

Hat Findings Schema

Every hat agent returns its findings as one structured object - the unit the chair synthesises, the critic reviews, and the discovery loop tests for new claims. It is the same shape in all three execution modes (solo, phased, team).

{
  "lens": "",
  "hat": "white|red|black|yellow|green",
  "claims": [
    { "claim": "", "severity_or_value": "HIGH|MEDIUM|LOW | positive | neutral", "evidence": "" }
  ]
}

• `lens` - the professional lens the finding came through (e.g. `security-eng`); empty/`blue` for solo, where the chair runs the hats directly.
• `severity_or_value` - reads by hat: **Black** uses risk severity (`HIGH`/`MEDIUM`/`LOW`); **Yellow**/**Green** use opportunity value (`positive`/`neutral`); **White** facts carry no severity (`neutral`); **Red** records the gut-check signal in the same field (e.g. `HIGH` unease, `positive` pull).
• `evidence` - the file, quote, datum, or reasoning the claim rests on. An empty `evidence` is what the completeness-critic flags as an unverified claim.

Capability Requirements

<!-- chat-replace:execution-mode-rule --> Three execution modes exist. Pick one **deterministically**: team size = 1 → **solo flat-parallel**; team size ≥ 2 AND you can confirm the team-mode capability (`SendMessage` plus background `Agent` teammates) is available → **team mode**; otherwise → **phased sub-agent mode**. Confirming availability means actively probing, not glancing at your visible tools - `SendMessage` may be deferred behind `ToolSearch` (see the capability-detection step). If, after probing, you still cannot reach team mode, default to phased - it degrades gracefully, whereas attempting team mode without the capability fails loudly.

| Mode | When chosen | Mechanism | Cost (relative) | |------|-------------|-----------|----------------| | **Solo flat-parallel** | Size 1 | Hat agents fire in parallel via standard Agent tool; Blue Hat synthesises | 1× | | **Phased sub-agent** | Size 2+, no flag | Iterate phases sequentially; spawn N sub-agents per phase (one per lens), each with fresh context, briefed via a running synopsis Blue Hat maintains | 2-4× | <!-- chat-replace:team-mode-row --> | **Team mode** | Size 2+, flag enabled | Persistent background `Agent` teammates in one implicit team cross-talk via `SendMessage` across phases | 5-15× |

<!-- chat-skip:start --> **Agent Teams flag** enables `SendMessage` and the background-teammate mechanism (`Agent` spawned with `run_in_background: true`, joined into one implicit team). Enable in your environment:

export CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1

Without the flag, `/huddle` still runs multi-perspective deliberations via phased sub-agent mode - it just trades cross-talk between agents for a running synopsis Blue Hat maintains as the persistent memory. Quality drops a little, cost drops a lot.

**Deferred-tool caveat (the silent-fallback trap).** With the flag enabled, newer Claude Code builds do not list `SendMessage` in your live tool set - they defer it behind `ToolSearch`, surfacing only its name in a system-reminder. If you decide between team and phased mode by glancing at your visible tools, you will wrongly conclude team mode is unavailable and degrade to phased with no error. Resolve it at the capability-detection step below by loading the schema with `ToolSearch("select:SendMessage")` before you decide - a successful load is confirmation that team mode is reachable. (This build forms a **single implicit team**: you spawn named background `Agent` teammates and they join automatically, so `SendMessage` is the one capability that gates team mode.)

**Why enable team mode anyway:** persistent professional-lens agents talking to each other across phases produce noticeably deeper synthesis - disagreements get rebutted in real time, edge cases surface from cross-talk, and the verdict feels like real deliberation rather than serially-summarised opinions. Worth it for decisions where being wrong costs 100× more than the analysis: architecture choices, irreversible migrations, hiring calls, contractual commitments.

**One team per session.** This build allows exactly one implicit team per Claude Code session, and the main session is its permanent lead. A team-mode huddle claims that single team - so do not start a second team-mode skill (another huddle, a `/tm` marathon) in the *same* session: its teammates would land in the same team, sharing one task list and one mailbox with the huddle's hats. To run two team-mode workstreams at once (e.g. a huddle defining the next PRD while a marathon implements the current one), use a *separate* session - a second terminal, ideally its own worktree. Each session gets its own isolated team (`session-<id>`-named), lead, task list, and mailbox. <!-- chat-skip:end -->

<!-- chat-replace:capability-detection --> **Tell the user which mode you're