deep-reviewlisted

# Deep review (Workflow-driven) The heavyweight form of `risk-review` for **L3/L4** diffs. It reviews each risk dimension in parallel and **adversarially verifies every finding**, so only confirmed risks reach the report — no plausible-but-wrong noise. ## When to use - **Explicit invocation only**, **L3/L4 only.** For a routine diff use `risk-review` (single pass) — say so and stop. - Drives the `Workflow` tool (~15× tokens) — invoking it is the opt-in. - **If `Workflow` is unavailable**, fall back to `risk-review`'s single pass and say so. ## How to run 1. Ensure there is a diff (`git diff`) and, if a spec exists, its acceptance-criterion IDs. 2. Run the Workflow in `${CLAUDE_SKILL_DIR}/workflow.md`: per risk dimension, find → adversarially verify each finding (≥2 skeptics, refute-by-default) → keep only confirmed; plus a `conformance-reviewer` pass on the diff vs the spec criteria. `log()` the dimension count. 3. Present **confirmed** risks (ranked, with file:line) · required approvals · missing / after-the-fact tests · suggested fixes · spec-conformance gaps. ## Output Ranked confirmed risks, each survived adversarial verification, plus spec-conformance gaps. Workers run via `agentType: "groundwork:<agent>"`. Reference script: `${CLAUDE_SKILL_DIR}/workflow.md`.