code-reviewer¶
Read-only review of a ticket's implementation diff against the approved plan, architecture invariants, and project conventions.
| Triggered by | /ticket:pick step 5.5, every loop round — one of the default blocking checkers via review.agents (default: [code-reviewer, test-adequacy-reviewer]) |
| Also usable | Standalone, on any uncommitted or branch diff |
| Tools | Read, Grep, Glob, Bash |
Flow¶
flowchart LR
In1["Ticket + plan"] --> A["code-reviewer"]
In2["Diff base<br/>(git diff --stat, then full hunks)"] --> A
In3["references.architecture /<br/>references.conventions"] --> A
In4["Prior-round findings<br/>(on re-review)"] --> A
A --> Out["Verdict + Blocking / Suggestions lists<br/>(file:line cited) + plan-fidelity note"]
Out --> Eval["Step 5.7 evaluate<br/>(re-plan / escalate / iterate / done)"]
Verdicts¶
| Verdict | Meaning |
|---|---|
PASS |
Clean — no findings |
PASS WITH SUGGESTIONS |
Non-blocking suggestions only |
BLOCKED |
At least one blocking finding |
Findings are capped at 10 blocking + 10 suggestions, each cited by file:line.
Gates directly? No — a blocking finding feeds the step 5.7 evaluation's decision (re-plan / escalate / iterate / done); the agent itself never edits or blocks.
See also¶
test-adequacy-reviewer— the other default blocking checker, focused on test quality rather than implementation quality