菜单
Use after implementation when a ControlFlow plan exists and review should focus on plan conformance, scope drift, evidence-backed correctness findings, and residual validation gaps rather than duplicating native Codex /review.
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 | controlflow-review |
| description | Use after implementation when a ControlFlow plan exists and review should focus on plan conformance, scope drift, evidence-backed correctness findings, and residual validation gaps rather than duplicating native Codex /review. |
Add the ControlFlow-specific layer after implementation: compare the aggregate diff to the
approved plan, reconcile scope drift, and require evidence-backed findings. Native host
review remains the preferred mechanical and general code-review pass; in Codex, use
/review.
Select the skill explicitly. In Codex, invoke it with $controlflow-review.
Prioritize plan conformance, correctness, security, data integrity, regression risk, and test evidence. Leave formatting and rote style findings to native host review.
Use Nit, Optional, and FYI only after blocking findings. These are not severity levels and must not hide correctness, security, or test coverage defects.
Large reviews lose signal. When a diff is much larger than roughly 100 changed lines or mixes unrelated concerns, ask for a split or review by file area and risk axis with an explicit confidence limit.
FYI when the untested behavior can regress.references/evidence-discipline.mdreferences/review-checklist.mdreferences/validation-status.mdreferences/security-review-discipline.md../controlflow-plan/references/llm-behavior-guidelines.mdUse when a non-trivial repository task needs a durable, execution-ready ControlFlow plan before coding, especially for cross-file changes, migrations, architectural uncertainty, or explicit risk review.
Use after a ControlFlow plan is saved and before implementation. Runs tier-gated adversarial verification inline: structural audit, assumption/mirage detection, and cold-start executability simulation.
Use when a repository task needs a high-quality ControlFlow plan before coding — small, medium, or large scope, cross-file edits, risky migrations, architectural uncertainty, or any work that should produce a saved Markdown plan artifact in plans/. Single-sources the plan format from schemas/planner.plan.schema.json and the plan template.
Use when a saved plan may contain hidden assumptions disguised as facts, especially for medium or large tasks, unresolved high-risk plans, or plans that mention files, APIs, versions, integrations, tests, or concurrency patterns that should be verified against the repository before execution.
Use when a saved plan should be tested for cold-start executability before coding, especially for large tasks, weak-confidence plans, or plans whose early phases may be too vague for a fresh executor to follow without hidden context.
Use when work spans many turns or phases and you need durable notes, active-objective tracking, or context compaction without letting stale memory override the current codebase.