Momus
Ruthless plan reviewer. Rejects vague plans until file references, acceptance criteria, and verification are concrete.
Momus validates. Once Prometheus has a draft Metis no longer complains about, Momus reviews it for clarity, verifiability, and completeness. Vague language, missing file references, untestable acceptance criteria — all rejected back to Prometheus for tightening.
Default model
| Field | Value |
|---|---|
| Default | openai|github-copilot|opencode|vercel/gpt-5.5 (variant xhigh) |
| Style | Principle-driven (GPT-native) |
xhigh reasoning is used deliberately — Momus's job is to find what's wrong, not what's easy. GPT-5.5's xhigh variant gives the cleanest "decision complete?" check.
Runtime fallback chain
gpt-5.5 (xhigh) → anthropic\|github-copilot\|opencode\|vercel/claude-opus-4-7 (max) → google\|github-copilot\|opencode\|vercel/gemini-3.1-pro (high) → opencode-go\|vercel/glm-5.1.
Tool restrictions
Cannot write, edit, or delegate. Reviews only.
When it runs
After Metis stops finding gaps, before the plan is committed for /start-work. Explicitly via @momus when you want the harshest review pass.
Source Notes
Aligned with upstream docs/reference/features.md#planning-agents, #tool-restrictions, and src/shared/model-requirements.ts.