codex-plan-reviewlisted

# Codex Plan Review Independent plan review using Codex (`gpt-5.6`, `model_reasoning_effort=xhigh`). This is for plans, roadmaps, specs, and design docs, not ordinary code diffs. > Requires the `codex` CLI (`npm install -g @openai/codex`, then `codex login`). > Persona listing/resolution works without it. ## Focus The generic review checks: - spec fidelity: missing requirements, invented scope, ambiguous acceptance criteria - internal consistency: formulas, thresholds, score ranges, naming, and definitions - dependency ordering: milestones that depend on unbuilt or underspecified pieces - data contracts: fields, enums, units, return shapes, schemas, and file paths - feasibility: work estimates, validation budgets, operational assumptions - edge cases: empty inputs, boundaries, NaN/None, rollback and failure modes - test adequacy: whether tests cover the actual risks in the plan ## Personas Personas are optional lenses, not a fixed enum. Decide which to use up front when the review is initiated. Example lenses: `architect`, `cto`, `red-team`, `software-architect`, `quant` — or pass a direct path to a persona `.md` file. Run without personas for a generic review; pass `--personas a,b,/path/to/persona.md` for a panel. List what's available via the shared resolver (works even without Codex): ```bash python3 ~/.claude/skills/persona-registry/scripts/persona_registry.py --list ``` ## Model & reasoning effort Defaults: **model `gpt-5.6`**, **reasoning effort `xhigh`**. T