From s2s
Facilitates multi-agent roundtable discussions using strategies including round-robin, Disney creative method, structured debate, consensus-driven, and Six Thinking Hats. Auto-detects approach from topic keywords.
How this skill is triggered — by the user, by Claude, or both
Slash command
/s2s:roundtable-strategiesThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
This skill provides facilitation strategies for s2s Roundtable sessions. Each strategy defines how participants interact, what phases the discussion follows, and how consensus is reached.
This skill provides facilitation strategies for s2s Roundtable sessions. Each strategy defines how participants interact, what phases the discussion follows, and how consensus is reached.
/s2s:roundtable session/s2s:specs requirements gathering/s2s:design architecture designWhen no strategy is specified, analyze the topic for keywords to recommend appropriate strategy:
| Keywords in Topic | Recommended Strategy | Reason |
|---|---|---|
| creative, innovation, new, brainstorm, ideation | disney | Creative ideation benefits from Dreamer→Realist→Critic flow |
| vs, compare, evaluate, choose, decide, between | debate | Comparing options needs Pro/Con analysis |
| urgent, fast, quick, asap, time-sensitive | consensus-driven | Speed-focused, converging quickly |
| comprehensive, thorough, all angles, deep, complete | six-hats | Complete analysis from all perspectives |
| (no match) | standard | Balanced default for most discussions |
Topic: "Creative approach for new authentication feature"
^^^^^^^^ ^^^
creative new
→ Recommended: disney strategy
| Strategy | Best For | Phases | Participation |
|---|---|---|---|
| standard | General discussions, balanced input | Single phase, multiple rounds | parallel |
| disney | Creative problem-solving, innovation | 3 phases: Dreamer, Realist, Critic | sequential (phases) |
| debate | Evaluating controversial options | 2 sides: Pro vs Con | parallel (per side) |
| consensus-driven | Fast decision-making | Converging rounds | parallel |
| six-hats | Comprehensive analysis | 6 thinking modes | sequential |
Added in v1.3.0 (TECH-002 Phase 4). Full worked examples + D3 hierarchy decision rationale:
${CLAUDE_PLUGIN_ROOT}/skills/roundtable-execution/references/strategy-resolution.md+ ADR-0011 Phase 4 addendum.
commands/roundtable.md Phase 1 resolves the active strategy in this fixed order. First source that yields a value wins; subsequent sources are NOT consulted.
CLI --strategy <value> (highest priority)
├─ if set → use it
└─ else → .s2s/config.yaml.roundtable.strategy.by_workflow_type[{workflow_type}]
├─ if set → use it
└─ else → profiles/{workflow_type}.yaml.default_strategy
├─ if set → use it
└─ else → error "no strategy resolvable" (defensive; should not trigger in normal use)
Forced override (wins over all): profiles/{workflow_type}.yaml.strategy_constraints.forced == true. Only brainstorm.yaml sets forced: true (forces disney regardless of CLI/config).
D3 source roles:
.s2s/config.yaml: user canonical (runtime authoritative, project-level overrides)profiles/{workflow}.yaml: plugin fallback (consulted only when config.yaml lacks the key)SKILL.md (table below): human-facing documentation only, NOT consumed at runtimeAuthoritative source: profile YAMLs in
${CLAUDE_PLUGIN_ROOT}/skills/roundtable-execution/profiles/{workflow}.yaml(fieldsdefault_strategy,participants.default,strategy_constraints). The table below is a human-readable summary; if it drifts, the YAML is correct.
Each workflow has a recommended default strategy and participant set:
| Workflow | Default Strategy | Default Participants | Rationale |
|---|---|---|---|
| specs | consensus-driven | PM, UX-Researcher, BA, QA | Requirements need broad agreement, not debate |
| design | debate | Architect, Security, TechLead, DevOps | Architecture trade-offs benefit from Pro/Con analysis |
| brainstorm | disney (forced) | Variable (--participants) | Creative exploration requires Dreamer→Realist→Critic |
| roundtable | standard | Architect, TechLead (configurable) | Generic discussion, single topic, no opinionated default |
Authoritative source: profile YAMLs
artifact_types[](fieldsprefix,session_key,is_primary). In-session artifacts are listed here.ADR-*(design) is a Phase 3 output artifact written to.s2s/decisions/, not an in-session entry insession.yaml.artifacts.
| Workflow | Primary Artifacts (in-session) | Secondary Artifacts (in-session) | Phase 3 output artifacts |
|---|---|---|---|
| specs | REQ-, BR-, NFR-* | OQ-, CONF-, EX-* | (none beyond session summary) |
| design | ARCH-, COMP-, INT-* | OQ-, CONF- | ADR-* (written to .s2s/decisions/) |
| brainstorm | IDEA-, RISK-, MIT-* | OQ-, CONF- | (none beyond session summary; ideas promoted to .s2s/ideas.md) |
| Strategy | specs | design | brainstorm |
|---|---|---|---|
| standard | ✓ | ✓ | - |
| consensus-driven | ✓ (default) | ✓ | - |
| debate | ✓ | ✓ (default) | - |
| disney | - | - | ✓ (forced) |
| six-hats | ✓ | ✓ | - |
Note: Brainstorm workflow ALWAYS uses disney strategy regardless of --strategy flag.
standard when:disney when:debate when:consensus-driven when:six-hats when:Strategies are loaded via config.yaml:
roundtable:
strategy: "disney" # Loads skills/roundtable-strategies/disney.md
# Optional overrides (warning if conflicts with strategy)
overrides:
participation: "sequential"
Each strategy file (standard.md, disney.md, etc.) defines:
defaults:
participation: "parallel" # How participants interact
phases:
- name: "phase-name"
prompt_suffix: "Additional instructions for this phase"
participants: "all" # or specific list
consensus:
policy: "weighted_majority"
threshold: 0.6
validation:
requires_sequential_phases: false
min_participants: 2
references/standard.md - Standard round-robin detailsreferences/disney.md - Disney Creative Strategy detailsreferences/debate.md - Structured debate detailsreferences/consensus-driven.md - Fast convergence detailsreferences/six-hats.md - De Bono's Six Hats details| Mode | Behavior | Mitigates |
|---|---|---|
parallel | All respond simultaneously | Sycophancy, echo chambers |
sequential | One at a time, sees previous | Enables building on ideas |
targeted | Facilitator picks who speaks | Focused expertise |
| Policy | Description |
|---|---|
simple_majority | >50% agreement |
weighted_majority | Confidence-weighted votes |
unanimous | All must agree |
When consensus fails after max_attempts, escalate to user with:
npx claudepluginhub spec2ship/spec2ship --plugin s2sConducts multi-round structured discussions among AI team roles (architect, UX, security) until consensus or max rounds. For converging diverse opinions into a single decision.
Executes multi-agent roundtable discussions for specs, design, brainstorm, and roundtable workflows. Triggered by /s2s commands or 'execute roundtable'.
Runs structured multi-perspective analysis using Six Thinking Hats with professional-lens agents. Scales from solo decision check to board-level deliberation.