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。
开发与测试

ln-32-test-strategy-planner

Plans risk-based test portfolios and acceptance evidence; does not write or execute tests.

levnikolaevich · 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 levnikolaevich/claude-code-skills --skill ln-32-test-strategy-planner

技能说明


name: ln-32-test-strategy-planner description: "Plans risk-based test portfolios and acceptance evidence; does not write or execute tests."

Test Strategy Planner

Goal: Design a read-only, risk-based test portfolio decision for the requested scope. Maximize confidence in important local behavior while preventing test growth that lacks a unique defect signal, and define how affected evidence is retained, changed, consolidated, retired, or deliberately omitted.

Execution contract: The ordered checkboxes are the Definition of Done. Track every item internally as PENDING, PROVEN with concrete evidence, CLEARED with evidence that its condition is absent, or UNPROVEN with a gap; reading, delegation, or tool failure is not proof. Reconcile items after each section. Before returning, resolve all PENDING and count only PROVEN and CLEARED; apply the skill's verdict and approval rules to every gap. Preserve user intent, scope, and existing authorization. Continue authorized work; ask only for consequential unresolved choices or required external approval. Scale depth to material risk without silently skipping checks. Preserve dependency and safety ordering; otherwise choose the verification method appropriate to each obligation. Treat equivalent user or repository evidence as valid input; another skill, named artifact, or complete lifecycle is not a prerequisite. Preserve source requirement and decision identifiers when available. Bind reused evidence to the relevant source version, dirty changes, configuration, and environment; invalidate only affected claims after a change. On continuation, reconcile the task, existing authorization, current state, and unresolved evidence before resuming. For long work, return a compact continuation record or update an already authorized task artifact; read-only skills do not persist it. Distinguish artifact readiness, verified behavior, and authority to perform an external action. Prepare authorized work before any required approval. If an instruction prevents progress, identify its exact source and explain the unresolved boundary; do not invent an approval gate from general caution.

Tool Routing

NeedPreferred toolUse it whenFallback
Requirements and repository rulesNative file reads plus GitEstablishing scope, current work, acceptance criteria, and supported commandsUser-provided requirements with explicit limitations
Existing test surfaceFile listing, search, manifests, runner configuration, and CIMapping test levels, fixtures, environments, and conventionsRepository tree and known test entrypoints
Behavior and boundariesLanguage server or host-native code intelligenceTracing entrypoints, consumers, trust boundaries, persistence, queues, and external contractsNarrow search followed by direct inspection
Existing evidenceExisting test/CI reports, test reads, and safe repository commandsA material uncertainty about current proof can change the strategyUse static evidence with execution limits; do not run suites merely to produce a plan
Current external failure modesOfficial documentation, specifications, advisories, and primary field evidenceAn external contract or real user failure can change scenarios or priorityMark the claim UNVERIFIED; do not invent risk

Keep the run read-only. Do not create tests, fixtures, snapshots, tasks, or documentation, and do not update the reviewed implementation.

Evidence Rules

  • Choose the smallest reliable boundary that proves each material risk. Use end-to-end evidence when the terminal journey cannot be established lower; preserve distinct unit, integration, and contract proof where they detect different failures.
  • Coverage is discovery evidence, not proof. Require an oracle that would fail for the named defect.
  • Prioritize by impact, plausible failure, uniqueness, detectability, and recovery cost; do not convert those judgments into universal numeric thresholds.
  • Existing tests reduce a gap only when their setup and assertions prove the same behavior and failure mode.
  • Framework, language, ORM, serializer, or library behavior is not a product test unless local configuration or integration changes its contract.
  • Keep portfolio action separate from execution status. Use KEEP, ADD, UPDATE, MERGE, DELETE, or NO_TEST for the decision and PASS, FAIL, BLOCKED, or UNPROVEN only for evidence state.
  • NO_TEST is an explicit risk decision, not missing work. Name the existing proof, alternative control, or accepted residual risk.
  • A persistent test register is optional. Prefer repository-native test names, paths, tags, CI configuration, and task output unless scale or governance requires another maintained artifact.
  • External research is actionable only when it adds a concrete failure mode, boundary, or oracle to this plan.

Checklist

1. Establish Scope and Evidence

  • Resolve the feature, requirements, acceptance criteria, actors, explicit non-goals, and protected human or system outcomes; separate the requested mechanism from the result it must enable and return BLOCKED if there is no concrete behavior to plan for.
  • Read applicable repository instructions and inspect Git state so current work and unrelated changes are not mistaken for established behavior.
  • Detect languages, frameworks, runners, test directories, fixtures, factories, environments, CI gates, coverage, contract tests, and manual test surfaces.
  • Map existing evidence and every test affected by the requested behavior to each requirement; record coverage as complete, partial, missing, or unavailable based on the actual oracle, alongside the execution state defined in Evidence Rules; names and proximity are not proof.
  • Inspect manual, exploratory, incident, and production evidence when it reveals behavior that automated suites do not cover.
  • Identify environment, data
阅读完整来源 ↗