[FEATURE] Per-skill model selection support

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Feature Request

Summary

Allow users to configure which model runs individual skills (slash commands), rather than applying a single model globally across the entire session.

Motivation

Different skills have very different complexity profiles:

• Heavy reasoning skills (/impl, /fix, /build) benefit from more capable models like Opus.
• Lightweight or repetitive skills (/status, /verify, /finish) work well on faster, cheaper models like Sonnet or Haiku.

Currently, the only way to switch models between skills is to manually run /model <name> before each invocation, which is error-prone and breaks flow. There is no way to encode model preference into the skill itself.

Proposed Solution

Support a model field in skill frontmatter:

---
name: impl
model: claude-opus-4-6
---

With a fallback chain: skill frontmatter → CLAUDE_CODE_SKILL_MODEL env var → session default.

Alternatives Considered

• Running /model manually before each skill — works but is tedious and easy to forget.
• CLAUDE_CODE_SUBAGENT_MODEL env var — only applies to subagents spawned inside skills, not the skill executor itself.

Use Case

▎ Set Sonnet 4.6 as global default for speed/cost, but have /impl and /fix automatically escalate to Opus for higher-quality output — without manual intervention.

Proposed Solution

No

Alternative Solutions

_No response_

Priority

Medium - Would be very helpful

Feature Category

CLI commands and flags

Use Case Example

_No response_

Additional Context

_No response_