review-plan

# Review Plan Run two AI plan reviews in parallel and return combined findings. ## Step 1: Identify the Plan Determine the plan to review: - If **plan text** is in conversation context, use it - If a **plan file path** was provided, read the file ## Step 2: Run Two Reviews in Parallel Launch two Agent tool calls in a single message so they run concurrently (`model: "opus"`, do not set `run_in_background`): ### Internal Plan Review Spawn a subagent with the full plan text and instruct it to: 1. Read project context (CLAUDE.md and files mentioned in the plan) to understand the codebase 2. Apply the plan determination criteria below 3. Return findings in the output format below ### Run `/peer-review` Skill Spawn a subagent whose prompt includes the full plan text and the following review prompt, and instructs it to invoke `/peer-review` via the Skill tool: ``` <task> Review the following implementation plan for issues that would cause an implementer to build the wrong thing or get stuck. Challenge the design direction: question whether the chosen approach is the simplest safe option and identify assumptions it depends on. </task> <dig_deeper_nudge> After surface-level issues, check for failure modes under stress: partial failure, race conditions, rollback safety, stale state, and data loss. </dig_deeper_nudge> <structured_output_contract> For each issue, state: (1) the problem, (2) where in the plan it occurs, (3) impact on implementation, (4) a suggested fix, and ...