feat: add /ai-cost skill — AI spend optimizer

[HMAKT99]

You're spending $3K/month on AI APIs. You could spend $800.

Most teams use Opus for everything when Haiku handles 80% of tasks. Identical prompts hit the API 100 times a day with no caching. System prompts carry 1,500 tokens that could be 400. Nobody tracks cost-per-feature.

What /ai-cost does

You:   /ai-cost

Claude: MODEL SELECTION AUDIT
        ═══════════════════════
        Location         Current Model     Task             Recommended    Savings
        chat.rb:45      claude-opus-4-6    Simple Q&A       haiku-4-5      ~90% ←
        summarize.ts    claude-opus-4-6    Summarization    sonnet-4-6     ~60%
        classify.py     claude-sonnet-4-6  Classification   haiku-4-5      ~80% ←

        Current: ~$3,200/month → After: ~$800/month (75% savings)

        CACHING OPPORTUNITIES:
        Classification   60% cache hit rate   → save $150/month (Redis, 1hr TTL)
        FAQ responses    80% cache hit rate   → save $200/month (Redis, 1hr TTL)

        COST PER FEATURE:
        AI Chat         $1,350/month   $0.135/user
        Auto-classify   $360/month     $0.036/user
        Summarization   $810/month     $0.081/user

Why this matters for gstack teams

Teams shipping 10K LOC/day are also shipping AI features. The AI API bill is the fastest-growing line item — and nobody has visibility into what's driving it. /ai-cost gives you the same structured analysis that /retro gives your engineering velocity.

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