| name | go-code-quality-review |
| description | Use for read-only review of newly written gocluster Go implementation when the work is High-risk or substantial: shared/exported interfaces, material algorithms or state machines, substantial rewrites, meaningful residual uncertainty, or multiple production packages with shared impact or cross-package uncertainty. Do not trigger for every Non-trivial Go edit. |
Go Code Quality Review
Overview
Use this skill after Go implementation work and before final closeout to review
the diff against GoCluster's code-quality, workflow, validation, and
operational standards. Use a separate non-steered reviewer when credible review
evidence depends on reasoning outside the implementation context. Otherwise the
method may remain lead-owned. The lead still owns fixes, validation claims,
traceability, and the final response.
Workflow
-
Confirm the phase boundary.
- Use after code is written for High-risk or substantial Go implementation.
- If a separate reviewer is used, keep it read-only and findings-only.
- Do not trigger this skill for documentation-only Markdown changes unless
the diff also changes Go code or a runtime/code contract.
-
Inspect the approved scope and current diff.
- Compare the diff to the approved Scope Ledger version and slice plan.
- Check for hidden scope expansion, speculative cleanup, compatibility
shims, fallback paths, future-proof hooks, or broad refactors.
- Name every changed Go file and any touched tests/docs that are material to
the review.
-
Review GoCluster code-quality standards.
- Apply
docs/code-quality.md, docs/change-workflow.md,
docs/review-checklist.md, and docs/dev-runbook.md.
- Check correctness, reviewability, missing error paths, no placeholders,
and comment intent.
- Check bounded retained state, eviction or deletion coupling, and
cardinality proof for maps, caches, interners, pools, indexes, or side
tables.
- Check goroutine, timer, channel, socket, queue, cancellation, deadline,
shutdown, and file-handle ownership where relevant.
- Check parser, protocol, config, YAML, operator-visible, and
support-agent impacts when relevant.
- Check hot-path allocation, CPU, lock, and p99 claims only against actual
benchmark/profile evidence.
-
Review validation evidence.
- Verify the selected validation lane matches the touched surface.
- Check whether targeted tests,
go test, go vet, staticcheck,
golangci-lint, race checks, fuzzing, benchmarks, or pprof were required
and reported.
- For command-backed concurrency, lifecycle, queue, timer, shutdown,
shared-state, or leak-detection claims, apply
docs/review-checklist.md
Command Evidence: check the command, scope, observed result, and the
rationale or minimal excerpt required for a high-risk claim. A bare
PASS/FAIL line is not sufficient for these claims.
- Do not generate or run transcript evidence for the lead agent. Report
missing, stale, skipped, failed, timed-out, partial, or waived evidence
accurately.
- Do not run final validation for the lead agent; identify missing or stale
evidence instead.
-
Report findings first.
- Order findings by severity.
- Include file paths and line references when inspected.
- Separate material findings, non-blocking observations, and remaining
unknowns.
Output Expectations
- Report material findings and inspected evidence without a mandatory heading,
taxonomy, or result envelope.
- The lead agent must disposition findings, make any fixes within approved
scope, rerun required checks, and preserve lead-owned closeout.