agentleFS
Sign inStart free
ZachArticulateVZachArticulateV/designer-pro-and-seo/skills/route-three-brain/SKILL.md

route-three-brain

Skill10 starsChanged 27 days agoMITView on GitHub

The routing law for handing off between Claude (primary driver), Codex/GPT-5.5 (adversarial review, second opinions), and Gemini (long-context, repo-wide synthesis). Codifies when each model is the right tool and enforces the no-self-review rule. Trigger when the user says "three brain", "route this", "which model should", "delegate to codex / gemini", or when a substantial Claude output needs an independent check.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If the user says "check your work" / "review this" / "is this right" / "second opinion" / "sanity check," route to Codex via …

    line 22
  2. Very specific:

    Startup self-check (once per session): Verify the tools exist: codex --version and gemini --version.

    line 37

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

route-three-brain

Family: routing Status: Stable

Purpose

The decision layer for cross-model routing. Three models, three roles:

  • Claude — primary driver for code, content, design generation, conversation.
  • Codex / GPT-5.5 — adversarial review, second-opinion debugging, deep correctness/security checking.
  • Gemini — large-context whole-repo analysis, multi-document synthesis, big structured-data passes.

Hard rule: Claude never reviews Claude — same architecture, same blind spots. If the user says "check your work" / "review this" / "is this right" / "second opinion" / "sanity check," route to Codex via route-codex-review.

Triggers

  • "three brain" / "tri-brain" / "route this" / "which model"
  • "delegate to codex" / "delegate to gemini" / "best model for X"

Inputs

  • Task description; the current output (if asking for review/handoff)
  • Constraints (cost, time, privacy)

Steps

  1. Startup self-check (once per session). Verify the tools exist: codex --version and gemini --version. If one is missing, announce it once and route only to what's available.
  2. Classify the task:
    • Build/generate/converse → Claude keeps it.
    • Review / verify / debug a substantial output, or any risk-path edit (auth, billing, migrations, deploy, secrets) → Codex via route-codex-review.
    • Needs more context than one window (whole repo, many docs) → Gemini via route-gemini-context.
  3. Announce forced routes in one line before running (so the user can interrupt): [three-brain] routing to Codex (review) — risk path: src/auth/.
  4. Execute the handoff via the execution skills and integrate the result — diff where models agree/disagree and adjudicate by evidence, not by averaging.

Outputs

  • A routing decision with justification
  • The integrated result of any executed handoff

Dependencies

  • route-codex-review (Codex execution), route-gemini-context (Gemini execution)
  • The codex and gemini CLIs installed (or the adjacent openai-codex / cc-gemini-plugin tools); degrades to a recommendation when neither is present

Notes

The law layer; the two route-* skills are the execution layers. Under-firing is fine; over-firing breaks trust. When uncertain, recommend rather than auto-route.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free