← ClaudeAtlas

plan-reviewlisted

Adversarially review a plan, work order, spec, or agent brief before anything gets built. Use when the user wants a plan reviewed, stress-tested, audited, or countersigned, says "plan review" or "poke holes in this plan", or wants a pre-build check on an issue/brief/PRD.
ConnorGriffin/skills · ★ 20 · Code & Development · score 66
Install: claude install-skill ConnorGriffin/skills
# Plan review Review a plan the way a cold, skeptical engineer would — **before** a line of it is built. The premise: a couple of adversarial cycles on a plan is cheaper than fixing the built thing after the fact. The subject can be anything plan-shaped: a GitHub issue, an agent work order, a PRD, a design doc, a chat-message plan. If the user didn't point at one, ask what to review — don't guess. **Read-only reviewer.** A plan review never edits code, and the reviewer never independently edits the order. It produces objections and a verdict. The caller-owned exception is defined in [Mechanical fix in place](#mechanical-fix-in-place). ## Cold means cold The reviewer must have no stake in the plan. If this session authored or co-authored the plan (or is unsure), do not review it yourself — dispatch a separate cold reviewer per plan through the interface below. Self-review by the author reliably misses what a cold reader catches, no matter how honestly the author tries to re-derive. ### Evidence-block spot check During the cold read, spot-check at least one `command → output` evidence block: run its recorded command and compare the complete result to the cited output. Treat a manual edit found in any evidence block as independently requiring a **BLOCKED** verdict. ## Delegation authority Invoking this skill authorizes every sub-agent dispatch that this procedure marks mandatory, including a mandatory nested review skill. Do not ask again solely because a session-level