bmad-build's description is too wide-ranging

[TeKRunneR]

Describe the bug
The description of the bmad-build skill is very wide-ranging, which causes coding agents to select it even when not appropriate.

Steps to reproduce
See screenshot below for an example where Codex triggered bmad-build in an inappropriate context. I just asked it to commit BMAD-installed files, and it instead started a review sub-agent because it tried to follow bmad-build's workflow. This happened with my second message after updating BMAD to 6.11.0 on the repo.

For what it's worth, after I asked Codex to explain why it thought that invoking bmad-build here was appropriate, its response was:

• My faulty reasoning was:

1. I saw “commit” as a repository mutation.

2. I overgeneralized the bmad-build description—especially “any … change request”—to mean any operation changing repository state.

3. I failed to distinguish:

   • changing repository content: build/fix/refactor → bmad-build
   • recording existing content: stage/commit/tag/push → direct Git workflow

4. After invoking it, I followed its rendered workflow mechanically instead of recognizing that reviews, specs, and tests were evidence the skill did not fit the
   request.

Expected behavior
bmad-build's description should make it clear that it should only be invoked when doing BMAD development. Other non-BMAD tasks should not result in it being invoked, at least not without consulting the user.

Environment (if relevant)

• Model(s) used: gpt-5.6-sol
• Agentic IDE used: Codex
• BMad version: 6.11.0
• Project language:

Additional context
I'm currently working around this by disabling implicit invocation of bmad-build by codex, but this shouldn't be necessary.

[coderabbitai]

🔗 Related PRs

#2703 - fix(project-context): admit commands whose obvious invocation is wrong [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!