[Feature] Configure model and effort per agent/workflow for Claude Code

[dominikaveresch-cc]

Problem

BMAD workflows have very different needs. Analyst, PM and Architect benefit from a
strong reasoning model with high effort, while story implementation, small fixes or
sprint planning run fine (and much cheaper/faster) on a smaller model with low effort.

Claude Code already supports this natively: a skill's SKILL.md frontmatter accepts
model (e.g. opus, sonnet, haiku, or a full model ID) and effort
(low | medium | high | xhigh | max). With context: fork, model also sets
the subagent's model.

Today there is no way to set these through BMAD:

• The TOML overrides in _bmad/custom/ cover persona, menu, facts and workflow
  behavior, but not model or effort.
• Editing the generated SKILL.md files in .claude/skills/ works, but every
  npx bmad-method install / update regenerates them and the changes are lost.
• Wrapper skills are a workaround, but they duplicate the command surface.

Proposal

Add optional model and effort keys to the customization surface, e.g.:

# _bmad/custom/bmad-agent-architect.toml
[agent.runtime]
model = "opus"
effort = "high"

# _bmad/custom/config.toml (team defaults per phase)
[runtime.defaults]
planning = { model = "opus",   effort = "high" }
build    = { model = "sonnet", effort = "medium" }
review   = { model = "sonnet", effort = "high" }

When installing for Claude Code, the installer writes these into the frontmatter of
the generated SKILL.md. Tools without such frontmatter support ignore the keys.
Precedence could follow the existing resolver: shipped → team → user.

Why this helps

• Lower cost and latency for routine Phase 4 work
• Stronger results where it matters (architecture, PRD, adversarial review)
• Settings survive updates and can be committed for the whole team

Related

• Claude Code skill frontmatter reference: https://code.claude.com/docs/en/skills