← ClaudeAtlas

review-planlisted

Review an implementation plan for completeness, correctness, over-engineering, and test coverage. Iterates review rounds until no critical issues remain or round limit hit. Activates on "review plan", "check the plan", "critique this plan", or as an optional step after planning:plan.
parMaster/claude-dlc · ★ 0 · AI & Automation · score 75
Install: claude install-skill parMaster/claude-dlc
# Plan Review Iterative structured critique of an implementation plan. A read-only review agent finds issues; the main session presents them and applies fixes on approval. ## Step 0: Find the plan file 1. If `$ARGUMENTS` contains a file path, use it 2. Otherwise check `docs/plans/` — most recently modified `.md` (excluding `completed/` and `wbs-*.md`) 3. If multiple plans exist and it's unclear which, list them and ask ## Step 0.3: Choose review runtime Before running any review round, check availability: `[ "$AGTERM_ENABLED" = "1" ] && command -v agtermctl >/dev/null 2>&1`. If unavailable, skip this step entirely and continue to Step 0.5 — there's only one real choice (this session), so there's nothing to ask. If available, ask with AskUserQuestion: ```json { "questions": [{ "question": "Where should this review run?", "header": "Runtime", "options": [ {"label": "This session", "description": "Continue here — delegates each review round to the plan-review subagent, as it already does"}, {"label": "Spawn Codex session", "description": "Hand the whole review off to a freshly spawned Codex CLI session running this same skill there — this session's job ends once it's spawned"} ], "multiSelect": false }] } ``` **This session**: continue to Step 0.5. **Spawn Codex session**: run `bash "${CLAUDE_PLUGIN_ROOT}/scripts/codex-review-handoff.sh" "PLAN_FILE"` (substitute the real plan path for `PLAN_FILE`). No model question — the spawned sess