AgentSkillsCN

analyzing-regression-scope

分析差异、变更文件、紧急修复及发布候选，以确定回归风险的扩散范围以及必须优先重新测试的部分。适用于变更后重新测试的范围界定、审查拉取请求的QA影响，或构建用于发布验证的最小信心测试套。

SKILL.md
--- frontmatter
name: analyzing-regression-scope
description: 'Analyzes diffs, changed files, hotfixes, and release candidates to identify where regression risk spreads and what must be retested first. Use when scoping retest after a change, reviewing QA impact for a pull request, or building a minimal confidence suite for release validation.'
argument-hint: 'Diff, changed files, PR summary, release notes, critical flows, and any known risk areas'
user-invocable: true

Analyzing Regression Scope

Use this skill when the real question is not "how do we test everything?" but "what must we retest now, and why?" It is built for risk-aware retest planning after code changes, configuration shifts, dependency updates, or release packaging.

When to Use

  • •analyze regression scope after a bug fix or feature change
  • •decide what must be retested for a hotfix
  • •prepare a release confidence suite from a diff or PR
  • •explain QA impact of shared component or config changes
  • •separate must-retest areas from nice-to-have regression coverage

Operating Principles

  • •Change surface first - understand what changed before proposing retest coverage.
  • •Risk spreads through dependencies - shared modules, auth, data mapping, and configuration often widen impact.
  • •Direct is not the whole story - the changed file is only the start; adjacent behaviors matter too.
  • •Small believable packs beat fake completeness - a focused confidence suite is better than a giant list nobody will execute.
  • •Unknown impact stays visible - if propagation is unclear, log it as an open question instead of bluffing.

Workflow

Phase 0: Frame the input

Use the best available source:

  • •changed files
  • •a diff or PR summary
  • •release notes
  • •a hotfix description
  • •a list of touched components or endpoints

Capture any known critical journeys, environments, or roles that matter.

Phase 1: Map the change surface

Break the change into four layers:

  1. •Directly changed behavior - the feature, endpoint, or component that was edited
  2. •Adjacent behavior - nearby flows that call, render, or depend on the changed surface
  3. •Shared risk - utilities, configuration, auth, layout, data mapping, or common services reused elsewhere
  4. •State and data risk - migrations, caching, permissions, synchronization, or environment-specific behavior

Phase 2: Classify retest risk

Use a practical three-level scale:

RiskUse when
HighDirect change on a critical path, shared core module, data integrity flow, auth, checkout, notifications, or high-traffic workflow
MediumIndirect dependency, shared helper, UI shell, validation logic, or meaningful edge-case exposure
LowIsolated change with narrow blast radius and no obvious shared-path impact

Phase 3: Build the regression pack

Organize the result into four buckets:

  • •Must retest - confidence blockers; skipping these would be reckless
  • •Should retest - meaningful adjacent coverage if time allows
  • •Can sample - smoke-level checks instead of full retest
  • •Can defer with caution - lower-impact items worth tracking, not necessarily executing now

For every important area, include:

  • •why it is affected
  • •what to retest
  • •what data or environment setup is needed
  • •whether the coverage should stay manual or move into automation later

Phase 4: Produce the report

Use ./resources/regression-scope-template.md when possible.

A strong report includes:

  • •a short change summary
  • •a risk-ranked regression table
  • •a minimal confidence suite
  • •explicit open questions and blind spots

Common Failure Modes

  • •equating changed files with full regression scope
  • •producing a huge retest list with no priority
  • •ignoring shared components, config, or auth ripple effects
  • •forgetting data setup or role-specific behavior
  • •treating uncertainty as low risk without explanation

Resource Map

  • •./resources/regression-scope-template.md - structure for regression reports and confidence suites

Related Skills

  • •designing-functional-tests - when the regression result should expand into manual scenarios or test cases
  • •verifying-acceptance-criteria - when the question becomes whether the feature now meets its stated contract
  • •requirements-test-coverage-mapper - when broader traceability matters more than tactical retest scope
  • •code-review-advanced - when the regression assessment should be paired with code-level risk review

Definition of Done

This skill is complete when:

  • •the change surface is explained clearly
  • •high, medium, and low regression areas are separated
  • •the must-retest pack is believable and prioritized
  • •open questions and blind spots are explicit
  • •the output can guide a tester or release lead immediately