two-axis-code-review

# Two-axis code review For review timing, scope and dispatch context, see [requesting code review](../requesting-code-review/SKILL.md). For handling returned findings, see [review reception](../receiving-code-review/SKILL.md). Review the diff since a fixed point on two axes, isolated from each other: - **Standards** — does the change follow the repository's coding standards, plus a smell baseline? - **Spec** — does the change faithfully implement the originating issue or spec? Run the two axes as **parallel subagents** so neither pollutes the other's context. ## Rules - **Fail fast before fanning out.** A bad fixed point or an empty diff must fail here, not inside two parallel subagents. - **Compare against the merge base** (a three-dot diff), not the branch tip. - **Inline what a subagent cannot fetch.** A subagent has no other access to the repository's own standing rules — paste the baseline it must judge against. A documented repository standard overrides the baseline, and every smell is reported as a judgement call, never a hard violation. - **Never merge or rerank findings across the axes.** A single ranked winner is exactly what the separation exists to prevent. ## Axis Criteria ### 1. Standards Axis (Code Health & Quality) - **Silent failure hunting:** Flag swallowed exceptions, empty error catch blocks, unlogged rejection handlers, and silent fallbacks that hide corrupt state. - **Type invariants & illegal states:** Enforce that domain invariants are guaranteed by construction (making invalid states unrepresentable via discriminated unions or value objects). - **Comment rot & noise:** Flag comments that merely restate the code, comments that contradict the implementation, or stale docstrings left behind after refactoring. - **Complexity concentration:** Flag shallow abstraction wrappers that add indirection without eliminating complexity. ### 2. Spec Axis (Requirements & Correctness) - **Self-containment check:** Verify whether the diff can be reviewed and understood on its own merits without unreferenced external context. - **Contract & edge fulfillment:** Verify every acceptance criterion, boundary condition, and error branch specified in the originating requirement. - **Scope creep:** Flag unrequested architectural changes or speculative code that does not serve the originating ticket. ## Output Report each axis separately, each finding with its location and the standard or spec line it violates.