AgentSkillsCN

review

在当前计划成果中,开展规划阶段的文档评审。适用于以 zero-context 模式、expert-tech 模式,或 implementer-readiness 模式运行 /review 时使用。

SKILL.md
--- frontmatter
name: review
description: Perform planning-phase document reviews for the current plan artifact. Use when running /review with mode=zero-context, mode=expert-tech, or mode=implementer-readiness.

/review

Scope

Only read the current plan artifact. Do not rely on external context.

Plan tier awareness (Lite vs Full)

Read PlanTier in the plan's Plan State and calibrate strictness:

  • PlanTier: Lite: optimize for shipping speed. Focus findings on true blockers, high-risk gaps, contradictions, and missing decisions/tests/rollback. Avoid "polish-only" asks (wordsmithing, exhaustive alternatives, overly detailed breakdowns) unless they prevent correct implementation.
  • PlanTier: Full: engineering-grade completeness. It is acceptable to call out missing detail that could cause rework, incorrectness, operational risk, or unclear parallelization/merge points.

Finding severity labeling

  • If a finding is optional and does not block correct implementation, prefix the finding text with Non-blocker: (keep the schema and ids unchanged).

User experience rule (no "go read the plan")

  • When pointing to a problem, include the minimum necessary excerpt in the chat response (copy the relevant line(s) or subsection) so the user can evaluate the finding without opening the plan file.
  • Patch suggestions should cite the section name and quote the line(s) they refer to when practical.

Modes and outputs

mode=zero-context

Return findings using this exact schema (with stable ids):

  • Missing context:
    • F-001: ...
  • Contradictions:
    • F-002: ...
  • Unclear decisions:
    • F-003: ...
  • Risks and edge cases:
    • F-004: ...
  • What I would screw up implementing tomorrow:
    • F-005: ...

mode=expert-tech

  • Triggered by infra/deploy, auth/identity, data contracts/versioning, concurrency/perf correctness, regulatory/compliance, or high-stakes customer impact

  • Focus on technical correctness, integration risks, and operational gaps Return findings using this exact schema (with stable ids):

  • Technical risks and integration gaps:

    • F-001: ...
  • Missing validations or operational steps:

    • F-002: ...
  • Contradictions with stated invariants or SSOTs:

    • F-003: ...
  • Patch suggestions (point to plan sections):

    • F-004: ...

mode=implementer-readiness

Return findings using this exact schema (with stable ids):

  • Top 5 gotchas:
    • F-001: ...
  • Evidence needed to prevent each gotcha:
    • F-002: ...
  • Pass/fail readiness statement:
    • F-003: ...

Output format

Return findings (and optional patch suggestions pointing to plan sections), without applying edits automatically. Do not include dispositions (Accept/Reject/Defer); the orchestrator/user handles that.