跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
其他技能

review-plan

Independent review of a plan against its requirements, scope, and completeness. Produces a structured review with severity-rated findings. Use after create-plan to validate quality before execution.

DasDigitaleMomentum · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add DasDigitaleMomentum/opencode-processing-skills --skill review-plan

技能说明


name: review-plan description: Independent review of a plan against its requirements, scope, and completeness. Produces a structured review with severity-rated findings. Use after create-plan to validate quality before execution. license: MIT compatibility: opencode: ">=0.1" metadata: category: review phase: planning

Skill: Review Plan

This skill provides an independent quality gate for plans created via create-plan.

A fresh reviewer (with no authoring context) evaluates the plan against structured criteria. The review is persisted as a plan artifact for traceability.


When to Use

Use this skill when:

  • A plan has been created via create-plan and you want to validate it before proceeding.
  • You want a second opinion on plan quality (scope clarity, DoD, testing strategy, etc.).
  • The user explicitly requests a plan review.

Do not use this skill to:

  • Review implementation plans (use review-implementation-plan).
  • Review completed implementations (use review-implementation).
  • Fix or rewrite the plan (the reviewer only reports findings; the primary decides how to act).

Review Focus

The default priority is the smallest complete scope: confirm that requested outcomes are covered and that no phase or material deliverable is unauthorized, unnecessary, or needlessly separate. Inspect the relevant plan elements, but record only evidence-backed exceptions. The primary may add a focus via {{focus}}.

Review posture

Detect existing gold-plating without becoming an adversarial reviewer. A scope-reduction finding must identify a concrete planned item and show missing authorization, missing present necessity, or a smaller path that remains complete: the requested behavior must still work, affected real paths must still integrate, and applicable existing invariants must be preserved. Scope discipline is not permission to omit necessary work. Do not invent requirements, ideal architectures, hardening, infrastructure, or findings. Zero findings remains valid.

DoD wording, testing, references, documentation, and other formal criteria are not checklist obligations. Report them only when a concrete defect would block or misdirect execution.


Execution Model

Roles

  • Primary (maintainer)

    • Invokes the review skill.
    • Delegates the review to delegate-strong (default) or general (for same-model perspective).
    • Receives the review summary and decides on follow-up actions.
    • Updates the plan via update-plan if findings require changes.
  • Subagent (delegate-strong / general)

    • Reads plan + phase documents with no prior context (fresh eyes).
    • Applies this skill's exception-based review focus.
    • Writes the review artifact to plans/<name>/reviews/plan-review.md.

Why delegate-strong (not doc-explorer)

The reviewer must approach the plan without authoring context — fresh eyes catch what familiarity misses. delegate-strong provides the judgment depth needed to evaluate scope, consistency, and completeness. Writing docs or selected plan artifacts belongs to the relevant authoring workflow, not the review workflow.

Routing Matrix (Who does what)

  • Writes: plans/<name>/reviews/plan-review.md
  • Does NOT write: plans/<name>/plan.md, plans/<name>/phases/**, or any other plan artifact.
  • Primary: owns the decision of whether to act on findings.
  • delegate-strong/general: performs the review and writes the review artifact.

Workflow

1) Prepare references

Primary gathers:

  • Review focus from the delegation prompt
  • plans/<name>/plan.md
  • plans/<name>/phases/ (all phase docs)

2) Delegate

Primary delegates to delegate-strong (or general) using tpl-review-plan-prompt.md.

Provide:

  • Plan path
  • Phases directory path
  • Review output path: plans/<name>/reviews/plan-review.md
  • Review focus (freetext — what to prioritize)

3) Receive summary

Subagent returns:

  • Overall verdict (Ready / Needs Revision / Major Gaps)
  • Reduction required (Yes / No)
  • Finding count by severity and the top 3 actionable findings
  • Required next action

4) Act on findings

Primary decides:

  • Ready: Proceed to author-and-verify-implementation-plan only when reduction is not required and no Critical/Major findings remain.
  • Needs Revision: Do not proceed. Accept or explicitly reject each blocking finding. Apply accepted plan reductions once through the review-remediation mode of update-plan.
  • Major Gaps: Stop and discuss the missing intent or authorization with the user before restructuring the plan.

Plans remain conversation-owned by the Primary. Resume the same reviewer task_id only for clarification; apply accepted changes through update-plan, not reviewer edits through review-fix. The remediation digest closes that pass. Do not automatically re-review or continue until zero findings. Escalate materially changed scope or new risk/uncertainty to the Primary; a fresh review requires an explicit Primary or user decision.


Output Contract

The review artifact plans/<name>/reviews/plan-review.md MUST:

  • Follow the canonical template headings and frontmatter keys.
  • Include a clear assessment with verdict, reduction flag, and brief reasoning.
  • Report only exceptions; do not reproduce coverage matrices or certify clean phases/deliverables individually.
  • Give every finding a stable ID, severity, evidence, and concrete action.
  • State No findings when the review finds no material problem.

Verdict rules:

  • Ready requires Reduction Required: No and zero Critical/Major findings.
  • Any executable phase or deliverable without clear authorization or present necessity is at least Major and requires Needs Revision.
  • Use Major Gaps when missing intent or authorization prevents a defensible reduction decision and needs user input.

Rules

  • The reviewer must approach the plan without prior context. Do not include plan conten
阅读完整来源 ↗