A structured review of a change across five axes — correctness, readability, architecture, security, performance — with findings pinned to file:line, categorized by severity (Critical / Important / Suggestion, or Nit / Optional / FYI), and paired with fix recommendations. The go/no-go signal a change needs before it merges.
Produced by (backlinks)
-
addy-code-review — the
/reviewskill workflow. -
addy-code-reviewer — the Staff-Engineer persona that emits the standard template (also in the
/shipfan-out). -
bmad-code-review — adversarial parallel review layers (Blind Hunter / Edge Case Hunter / Acceptance Auditor) with structured triage.
-
ce-code-review — Compound Engineering’s ~16-persona confidence-gated review (adversarial + cross-model).
-
ce-doc-review — Compound Engineering’s plan/requirements-side persona review.
-
gstack-review — gstack’s Staff-Engineer pre-landing PR review (auto-fixes the obvious ones).
-
gstack-codex — cross-model second-opinion review (OpenAI Codex).
-
gstack-design-review · gstack-devex-review · gstack-ios-design-review — live design / developer-experience / iOS-HIG audit reports.
-
mp-code-review — Matt Pocock’s two-axis review: Standards (+ Fowler smell baseline) ∥ Spec, run as parallel sub-agents.
pstack:
- pstack-interrogate — a synthesized verdict from one reviewer per model against a shared rubric; explicitly does not auto-apply changes.
- pstack-blast-radius — a what-it-breaks writeup whose central section is the one fact it is safe because of, graded on a five-rung evidence ladder and marked unproven below rung 4.
- pstack-comment-sicko — touched files, deletion count,
MUST KILLflags one line each, and skips. - pstack-no-comments — the action report: deletions, restorations, reruns, fixes, encoding offers, and unenforced constraints left open.
Tessl — review reports at three different subjects, which is the layer’s tell:
- tessl-review — scores a skill across validation, description, and content; runs automatically on publish and the score travels with the registry entry.
- tessl-review-security — severity-graded supply-chain findings on a skill or plugin (pattern-supply-chain-scanning).
- tessl-code-review — an agentic pull-request review with versioned, pinnable lenses.
- tessl-change-risk · tessl-change-verify — a risk verdict and per-file judge findings over a diff, both with CI exit codes.
- tessl-inventory — org-wide Findings, ranked by severity, across every repo scanned.
builder-skills:
- builder-factory-review-prs — per-PR findings with file, line, impact, and a concrete fix, kept separate from approval and merge decisions. A PR with no findings is recorded as such, with no comment invented.
- builder-agent-watchdog — the gap report: Status / Requested / Observed / Gaps / Independent findings / Fixes made / Remaining risk, classifying each issue as gap, bug, verification miss, scope drift, or no issue.
See Also
- artifact-security-audit · artifact-perf-audit — the security and performance reports that join it in the
/shipmerge. - stage-review — the stage this artifact gates.