Review the diff between HEAD and a fixed point the user supplies. Each axis runs as its own parallel sub-agent with its own brief under axes/, so no axis's findings colour another's. This skill selects the axes, dispatches them, and aggregates.

Axis	Question	Runs when
Scope	Can a reviewer take this in one pass, or should it be a stack?	>400 changed lines (excluding lockfiles and generated files), or >15 files, or the commits span more than one unrelated concern
Spec	Does it do what the spec asked?	a spec is found (step 2)
Correctness	Is it right, independent of the spec?	always
Security	Can untrusted input reach a dangerous operation?	the diff touches input parsing or request handlers, authn/authz, crypto, secrets or config, shell/SQL/path/URL construction, deserialization, or dependency manifests
Performance	Is it needlessly slow or costly at realistic scale (extra round-trips, missing index, N+1, unbounded work)?	the diff adds or changes code that performs I/O (database, network, file, cache), queries, loops over collections, or schema/index definitions
Standards	Is it well made, per the repo's standards and the smell baseline?	always
Testability	Can the production code be unit-tested cheaply through its interface?	the diff adds or changes production logic (not config, docs, or generated files)
Test Quality	Are the tests thorough, correctly labeled, and of high quality?	the diff touches production logic or tests

The brief for each axis is axes/<axis>.md (test-quality.md for Test Quality), relative to this skill's directory.

The issue tracker should have been provided to you. If docs/agents/issue-tracker.md is missing, tell the user to run /setup-matt-pocock-skills.

code-review

Process