plan-review-loop

What problem does it solve?

Hobo-specific operational playbook for the P3 Plan Review Loop — how to draft a plan that survives review, how to scaffold revision-prompts so round 2 is a verification pass instead of a re-read, when to STOP looping (severity trajectory + actionable-vs-zero exit), what plan-reviewer is structurally blind to (orthogonal axes, runtime collisions), and the same iterative pattern applied to codex review of prescriptive process docs. Load BEFORE invoking the plan-reviewer subagent in P3, BEFORE submitting a v2/v3 revision after addressing prior findings, BEFORE deciding whether a "no high-severity" round means done, and BEFORE running codex review --base main against changes to docs/process/, .claude/agents/, .claude/skills/, or any prescriptive workflow doc. Also load when the user says "plan-reviewer", "プランレビュー", "Plan Review Loop", "P3 のレビュー", "review loop", "何ラウンド回せばいい？", "もう収束した？", "actionable findings", or asks why plan-reviewer keeps surfacing new findings each round.

Core Features & Use Cases

The skill provides a repeatable, round-based protocol for plan reviews, including load-time pre-checks to ensure audience alignment; a convergence model based on actionable findings; scaffolding prompts to speed round-2 revisions; clear exit criteria when no actionable findings remain; and guidance on handling orthogonal runtime concerns to avoid blind spots.

Quick Start

Load this playbook before starting a plan review to scaffold round-2 prompts and accelerate convergence.