feat: add /ai-council skill — AI advisory board with 5 specialist perspectives

[HMAKT99]

One perspective misses tradeoffs. Five perspectives surface them.

Should you use Opus or Haiku? Build your own model or use the API? Add content filtering that adds 200ms latency? A single perspective gives you one answer. Five specialists debating gives you the tradeoffs — and the tradeoffs are what you actually need to make a good decision.

What /ai-council does

You:   /ai-council --decision "Should we use AI for search ranking?"

Claude: COUNCIL DELIBERATION
        ─────────────────────
        Evaluator:    "Current relevance: 72%. AI could hit 90%."
        Cost Analyst: "That's $800/month for 50K queries."
        Safety:       "Search ranking = bias risk. Need fairness testing."
        Architect:    "Latency budget: 200ms. LLM call: 500ms."
        Product:      "Users complain about search. #1 pain point."

        DEBATE:
        Cost vs Product: "$800/month for the #1 user complaint? Easy ROI."
        Architect vs Product: "Can't add 500ms. Use embeddings, not LLM."
        Safety vs all: "Add fairness eval before launch. Non-negotiable."

        RESOLUTION: Re-rank top 20 with cached embeddings.
        Latency: <100ms. Cost: ~$200/month. Fairness eval required.
        VOTE: Approved 5-0 with conditions

5 council members, distinct biases

Member	Cares about	Bias
Evaluator	Output quality, eval coverage	"Prove it with evals"
Safety Officer	Injection, PII, bias	"What's the worst case?"
Cost Analyst	Token costs, model selection	"Can we use Haiku?"
Architect	Latency, reliability, scale	"What happens at 100x?"
Product Owner	User value, time to ship	"Does this help users?"

Extends gstack's "explicit gears" philosophy

gstack already has CEO mode vs Eng Manager mode vs Staff Engineer mode for code decisions. /ai-council brings the same multi-perspective rigor to AI decisions.

Only .tmpl committed — bun run gen:skill-docs generates the rest.

Test plan

• .tmpl follows template pipeline — uses {{PREAMBLE}}
• Registered in gen-skill-docs.ts, skill-check.ts, both test files
• bun run gen:skill-docs generates valid SKILL.md
• All existing tests pass with skill added