Agent Skills: Convene Council

Convene Council

Convene four advisors for ambiguous decisions: the in-context Claude voice (Architect) plus Skeptic, Pragmatist, and Critic subagents (lenses in the Roles table). For decision-making under ambiguity only — see When NOT to Use.

When to Use

Use convene-council when a decision has multiple credible paths and no obvious winner, tradeoffs need explicit surfacing, the user asks for second opinions or dissent, or conversational anchoring is a real risk (e.g. monorepo vs polyrepo, ship now vs hold, feature flag vs full rollout).

When NOT to Use

| Instead of convene-council | Use | | --- | --- | | Verifying whether output is correct | review-with-doubt or review-blindly | | Breaking a feature into implementation steps | plan-tasks | | Reviewing code for bugs or security | review-blindly or review-with-doubt | | Validating completed work | review-work-completion | | Straight factual questions | just answer directly | | Obvious execution tasks | just do the task |

Roles

| Voice | Lens | | --- | --- | | Architect | correctness, maintainability, long-term implications | | Skeptic | premise challenge, assumption breaking, the simplest credible alternative | | Pragmatist | shipping speed, simplicity, user impact, real-world execution | | Critic | edge cases, downside risk, failure modes, reasons the plan could fail |

The three external voices should be launched as fresh subagents with only the question and relevant context, not the full ongoing conversation. That is the anti-anchoring mechanism.

Workflow

1. Extract the real question

Reduce the decision to one explicit prompt:

• what are we deciding?
• what constraints matter?
• what counts as success?

If the question is vague, ask one clarifying question before convening the council. Unattended (CLAUDE_UNATTENDED=1): skip the question — take the stated defaults, log defaulted:<decision> in the verdict, per ~/.claude/skills/run-autopilot/references/unattended-contract.md.

2. Gather only the necessary context

Codebase-specific decision: collect the relevant files, snippets, issue text, or metrics — compact, only what the decision needs. Strategic/general decision: skip repo snippets unless they materially change the answer.

3. Form the Architect position first

Before reading other voices, write down:

• your initial position
• the three strongest reasons for it
• the main risk in your preferred path

Do this first so the synthesis does not simply mirror the external voices.

4. Launch three independent voices in parallel

Launch them via the Agent tool (subagent_type: general-purpose) in a single message, in parallel — each gets only the prompt below (question, compact context, strict role; no conversation history, per the Roles-table rule).

Prompt shape:

You are the [ROLE] on a four-voice decision council.

Question:
[decision question]

Context:
[only the relevant snippets or constraints]

Respond with:
1. Position — 1-2 sentences
2. Reasoning — 3 concise bullets
3. Risk — biggest risk in your recommendation
4. Surprise — one thing the other voices may miss

Be direct. No hedging. Keep it under 300 words.

Each role's emphasis is its Roles-table lens — restate nothing else.

5. Synthesize with bias guardrails

You are both a participant and the synthesizer, so use these rules:

• do not dismiss an external view without explaining why
• if an external voice changed your recommendation, say so explicitly
• always include the strongest dissent, even if you reject it
• if two voices align against your initial position, treat that as a real signal
• keep the raw positions visible before the verdict

6. Present a compact verdict

Use this output shape:

## Council: [short decision title]

**Architect:** [1-2 sentence position]
[1 line on why]

**Skeptic:** [1-2 sentence position]
[1 line on why]

**Pragmatist:** [1-2 sentence position]
[1 line on why]

**Critic:** [1-2 sentence position]
[1 line on why]

### Verdict
- **Consensus:** [where they align]
- **Strongest dissent:** [most important disagreement]
- **Premise check:** [did the Skeptic challenge the question itself?]
- **Recommendation:** [the synthesized path]

Keep it scannable on a phone screen.

Persistence Rule

Do not write ad-hoc notes to shadow paths from this skill.

If the council materially changes the recommendation:

• update the relevant GitHub / Linear issue directly if the decision changes active execution truth
• save a memory file under the current project's auto-memory directory (~/.claude/projects/<project-hash>/memory/) only if the outcome is durable guidance for future sessions

Only persist a decision when it changes something real. Never persist every verdict regardless of importance.

Multi-Round Follow-up

Default is one round. On a requested second round: keep the new question focused, include the previous verdict only if necessary, and keep the Skeptic clean to preserve anti-anchoring value.

Related Skills

• review-with-doubt — adversarial self-review of a single output
• research — gather external reference material before the council if needed
• debug-stuck-agent — when the issue is a stuck agent rather than an ambiguous decision

The value is not unanimity. The value is making the disagreement legible before choosing.