workbench-code-review
Workbench Code Review
Use this skill for reviewing code, diffs, task plans, live workflow changes, and implementation evidence before merge or closure.
Review Stance
- Findings first. Do not lead with praise or a broad summary.
- Order findings by severity and user impact.
- Tie every finding to a file, command, issue/comment ID, run ID, or observable behavior.
- Prefer concrete repro or blast-radius evidence over style opinions.
- If no issues are found, say that clearly and name remaining test gaps.
Severity
P0: security, data loss, irreversible action, or broken primary workflow.P1: likely user-facing regression or unsafe automation.P2: correctness gap, missing edge case, or weak verification.P3: maintainability, clarity, or low-risk cleanup.
Checklist
- Does the implementation match the promised objective?
- Are owner boundaries preserved?
- Are live mutations backed up and reversible?
- Are commands, IDs, and paths reproducible?
- Are tests or real-path checks sufficient for the blast radius?
- Did the change avoid unrelated refactors and hidden policy changes?
Output Contract
For findings:
SEVERITYLOCATIONISSUEWHY IT MATTERSREQUIRED FIX
For a clean review:
VERDICT: PASSEVIDENCE CHECKEDTEST GAPSRESIDUAL RISK
More from fearvox/multica-ultimate-workbench
workbench-conductor
Two-ring orchestration, routing, issue and comment discipline, and role boundaries for the Multica Workbench.
4workbench-self-awareness-infra
Capability discovery and current-state verification for Heavy Path, ambiguous repo/runtime ownership, and runtime-dependent Standard Path work.
4workbench-sdd
Specification-driven development from raw requirement to product design, technical design, task list, execution, and verification.
4workbench-debug-investigate
Root-cause investigation for bugs, regressions, suspicious runtime behavior, failed automations, and unexpected outcomes.
3workbench-research
Evidence-first research, architecture discovery, vendor documentation review, benchmark reading, and unknown-system analysis.
3workbench-goal-mode
Goal-persistence execution wrapper for autonomous work that must continue across turns until the stated objective is actually satisfied.
3