菜单
Defines safe sandbox-to-production promotion gates for agent-built software. Use when deploying, restarting, promoting builds, validating runtime surfaces, checking rollback readiness, or proving that the running production artefact matches the reviewed source commit.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | sandbox-to-production-gates |
| description | Defines safe sandbox-to-production promotion gates for agent-built software. Use when deploying, restarting, promoting builds, validating runtime surfaces, checking rollback readiness, or proving that the running production artefact matches the reviewed source commit. |
Use this skill when an agent wants to move work from local or sandbox into production.
The goal is simple: prove that the reviewed source, built artefact, deployed service, and live runtime are the same thing.
Green tests are not a deployment. A restart is not a deployment. A deployed bundle without smoke evidence is not a deployment.
Capture:
If there is unrelated dirty work, do not stage it. Isolate the promotion.
Run the real build command for the promoted surface.
Capture:
Do not claim a build passed because a dev server still runs.
Identify the actual runtime service:
Restart only the intended service. Never restart a neighbouring prod service by matching a vague process name.
Smoke the live runtime, not localhost in the wrong mode.
Check:
If a public CDN or proxy blocks automation, smoke the backend with the correct Host header and say that is what you did.
Before promotion, know the rollback:
If rollback is not known, label the promotion high risk.
Return:
Commit promoted:
Build command:
Build result:
Service restarted:
Runtime smoke:
Logs checked:
Rollback path:
Dirty state:
Stop and ask or hold if:
Preserves useful state across long agent sessions, context compaction, handoffs, and restarts. Use when compressing context, handing work between agents, resuming after interruption, writing session summaries, or preventing repeated work.
Enforces a design/spec approval gate before production implementation. Use when UI, UX, product flows, architecture, API shape, prototypes, visual direction, or user-facing behaviour need signoff before builders write production code.
Keeps agent work safe in messy repositories. Use when git status is dirty, unrelated files exist, generated artefacts are present, multiple agents share a repo, or you must commit only the intended slice without damaging someone else's work.
Proves model and provider capability claims before changing agent defaults. Use when swapping models, checking context windows, reasoning effort, tool use, streaming, media support, OAuth entitlement, fallback behaviour, or provider-specific runtime paths.
Converts bug reports, screenshots, user panic, QA notes, and vague complaints into durable engineering briefs. Use when turning observed failures into reproducible issues with severity, expected behaviour, evidence, scope, owner, and acceptance tests.
Defines strict review-then-fix authority for reviewers. Use when a reviewer is allowed to repair findings after review, when avoiding reviewer-as-stealth-builder drift, or when verifying that fixes stay limited to documented findings.