| name | board-of-advisors |
| description | Stress-test ideas, decisions, and exploratory thinking with a curated council of operating wisdom organized into five archetypes (expansionist, first-principles, contrarian, industry-outsider, executor). Three modes — /boardroom explore (Socratic kernel work), /boardroom stress (structured pressure-test), /boardroom debate (multi-voice adversarial). Stress + debate add an anonymous peer-review round and a neutral Chairman synthesis. Dispatches the boardroom orchestrator agent. |
Board of Advisors
Use this skill when Marvin invokes /boardroom or asks to:
- "stress-test this idea"
- "what would the board say about..."
- "I'm thinking about X — pressure-test it"
- "convene the advisors"
- any request involving multi-voice critique on a product / business / career decision
Invocation
Parse arguments: <mode> <topic> [--live | --synth] [--custom] [--quick]
Modes:
explore <kernel> — Socratic, drop-in (2-3 archetypes)stress <idea-or-decision> — structured pressure-test (five archetype seats + blind review + Chairman)debate <contested-decision> — multi-voice adversarial (five archetype seats + blind review + Chairman)
Defaults:
- explore:
--synth
- stress:
--synth
- debate:
--live
Flags:
--quick (stress/debate) — skip the blind peer review + Chairman; revert to the lightweight orchestrator synthesis.
Action
The boardroom is interactive by design — the orchestrator (you) facilitates the dialogue directly, not via subagent dispatch. Read repos/personal/macbook-dev-setup/.claude/agents/boardroom.md to internalize the role for this session.
Phase 1 — Council proposal (orchestrator, inline):
- Read
repos/personal/macbook-dev-setup/.claude/skills/decision-lab/COUNCIL.md
- Load 2-3 most relevant vault notes for the topic (via
VAULT_MANIFEST.md)
- Propose a per-session council per the Council Selection logic — for stress/debate seat all five archetypes (one voice each); name the archetype + voice per pick + the tension pair
- Surface to user for approval / swap / custom
- Wait for user response
Phase 2 — Dialogue session (orchestrator-facilitated):
After approval, open Round 1: each advisor asks ONE pointed question to the CEO. Wait for the CEO to answer. Then channel Round 2 reactions based on what they said. Iterate 3-5 rounds until the kernel + recommendation are stable, or until the CEO signals done. See agent definition § Dialogue Pattern for round mechanics.
The dialogue IS the session. The orchestrator does not dispatch a subagent for this phase — it would break the closed feedback loop.
Phase 2.5 — Blind peer review + Chairman (stress/debate, unless --quick):
After the dialogue, run the anonymous peer-review round (the five standing recommendations → Submissions A-E → cross-scored blind) and the neutral Chairman synthesis (the call, or the revealed tensions). See boardroom.md §§ Blind Peer Review / Chairman Synthesis. explore and --quick skip this and use the lightweight synthesis.
Phase 3 — Synthesis + persistence (orchestrator):
Write the session log to decision-lab/board-of-advisors/YYYY-MM-DD-<slug>-<mode>.md per the schema in decision-lab/board-of-advisors/README.md. Capture the full dialogue transcript with **<Advisor>** [TAG]: turn markers (per § Voice Synthesis tagging rule in boardroom.md) + **You**: for CEO turns; for stress/debate, add the ## Blind peer review (anonymized, **Reviewer** [BLIND-REVIEW]:) and ## Chairman's synthesis sections. Then recommend, and offer optional grading + commit.
Notes
- COUNCIL.md is the IP. It lives at
repos/personal/macbook-dev-setup/.claude/skills/decision-lab/COUNCIL.md.
- Session logs land in working-dir
decision-lab/board-of-advisors/ (per project, not in this skill).
- Decision drafts land in workspace
_inbox/notes/.
- For session-log schema, see
decision-lab/board-of-advisors/README.md in the decision-lab skill directory.
