메뉴
Alias for /triple-review — drive a refactor through the full quad-review methodology (Codex + Antigravity + Claude SMR + Copilot) with plan-review, smoke, and 4-of-4 merge gate.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Drive ALL open issues to terminal (MERGED / PLAN-KILLED / PLAN-DEFERRED-with-reason) — a drive-to-zero campaign. Ranks the backlog, fans out /research + /engineer agents (parallel route) or chains coupled work (serial route), PARENT-reviews every PR, files new issues found during audits, and stops when nothing driveable remains.
Merge all open PRs (or specific ones) and close their associated issues
Run IPv4/IPv6 throughput and CPU profiling on bpfrx cluster or standalone VM using iperf3 + perf record.
Research an issue with hostile reviews from Claude SMR + Codex + AGY; produce a plan-of-action doc for manual approval BEFORE any implementation. Stops at PLAN-READY — does NOT engineer the PR.
Drive a hostile network-expert quad-review on a GitHub PR — Claude hostile in-conversation, Codex hostile, Antigravity adversarial, plus Copilot inline reviews. All four reviewers hostile-verify rather than confirm. Iterate across force-pushes until all reviewers agree. Never autonomously merge — synthesis only.
Verify the core security/forwarding matrix on real traffic — plain forwarding at max throughput, trust→untrust ALLOW, untrust→trust BLOCK, trust→trust ALLOW — with captured proof, then document the evidence.
| name | engineer |
| description | Alias for /triple-review — drive a refactor through the full quad-review methodology (Codex + Antigravity + Claude SMR + Copilot) with plan-review, smoke, and 4-of-4 merge gate. |
| user_invocable | true |
/engineer <issue-number> <one-line scope> is identical to
/triple-review <issue-number> <one-line scope>. Both invocations
run the same quad-review methodology codified in
.claude/skills/triple-review/SKILL.md.
When invoked, immediately invoke the triple-review skill via
the Skill tool, forwarding the arguments verbatim:
Skill({skill: "triple-review", args: "<verbatim args>"})
Do not duplicate the methodology here. The single source of
truth is .claude/skills/triple-review/SKILL.md — any change
to the methodology (e.g. the Claude SMR gates at Step 3.5 /
Step 8.5, the marker-post protocol, the smoke-runner handoff)
applies to /engineer automatically because this skill is a
thin forwarder.