Add subagent model-routing role agents to onboarding#56
Merged
Conversation
Add four Claude Code subagent role definitions (onboarding/agents/) and a "Subagent model routing" rule so any-model mother sessions dispatch to a role agent that pins its own model + effort, instead of ad hoc per-task choices: - snippet-classifier -> haiku (trivial mechanical work) - code-implementer -> sonnet (default well-scoped implementation) - architecture-reviewer -> opus (design review / strategy) - project-executor -> fable (large single-session missions) The routing is model-agnostic: each role pins its model in frontmatter, so a Sonnet/Opus/Fable/Haiku mother that dispatches code-implementer still gets Sonnet. Verified end-to-end: a Fable-5 headless session dispatching code-implementer ran the subagent on claude-sonnet-5. CLAUDE.global.md gains the routing section under the existing subagent preference; ONBOARDING step 8 symlinks the agents into ~/.claude/agents/; MAINTAINING records the files and their lockstep with the routing section. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds a model-routing layer for Claude Code subagents to the group onboarding, so a mother session (running on any model) dispatches work to a named role agent that pins its own model +
effort, instead of making an ad hoc per-task model choice.Four role agents in
onboarding/agents/:snippet-classifiercode-implementerarchitecture-reviewerproject-executorWhy this shape
code-implementerstill gets Sonnet.code-implementervsproject-executor— is written on the scope/autonomy axis, defaulting tocode-implementerwhen unsure.project-executoris one dispatch (no/handoff, no Phoenix, no context loop); the docs point at theautodevskill for genuinely unbounded work.Changes
onboarding/agents/*.md— the four role definitions (model+effortfrontmatter; both officially supported fields, verified against Claude Code sub-agent docs).onboarding/CLAUDE.global.md— new "Subagent model routing" section under the existing subagent-preference rule.onboarding/ONBOARDING.mdstep 8 — symlink-install into~/.claude/agents/(symlinks sogit pullkeeps them current) + a "takes effect next session" caveat.onboarding/MAINTAINING.md— provenance row; agent files and theCLAUDE.global.mdrouting section must stay in lockstep.Verification
End-to-end, not just self-report: a fresh Fable-5 headless session dispatching
subagent_type: code-implementerran the subagent onclaude-sonnet-5(read from the actual stream, not model introspection) — confirming the frontmatter pins the model and that routing is model-agnostic.🤖 Generated with Claude Code