analyze-ci-failures

作者: microsoft

分析 vcpkg Azure DevOps CI 失败原因。下载日志,识别回归根本原因,按包和三元组生成报告。用于:CI 失败…

npx skills add https://github.com/microsoft/vcpkg --skill analyze-ci-failures

vcpkg CI Failures Analyzer

Overview

Consumes already-collected CI evidence, cross-references it with baselines, and produces a regression report. If evidence has not yet been collected, invoke fetch-vcpkg-ci-logs first, then return to this workflow. Do not duplicate its acquisition steps here.

MCP Tools

ToolPurpose
github-mcp-server-get_file_contentsRead baseline files

Prerequisites

  • Local failed step logs and relevant extracted failure-log artifacts.
  • Build metadata and full Azure build URL.
  • For PR builds, PR metadata and full PR URL.

Evidence Acquisition

When the required evidence is not already local, delegate acquisition to a subagent using fetch-vcpkg-ci-logs. Prefer a fast, low-cost model for this mechanical work. Ask the subagent to return only the evidence inventory and local paths; keep regression analysis and report generation in the main agent.

When the evidence is already available locally, proceed directly to analysis.

Workflow

OUTPUT RULE: This skill applies only when the user asked for triage or report output, so a report.md is always required once it is selected. Write it as soon as step-log analysis completes, then enhance it from artifacts. Your response MUST also contain the complete report content — not a summary.

URL RULE: Your response text (not just the report file) MUST include:

  • For scheduled/manual builds: the full Azure DevOps URL https://dev.azure.com/vcpkg/public/_build/results?buildId={buildId}
  • For PR builds: both the full PR URL https://github.com/microsoft/vcpkg/pull/{prNumber} AND the Azure DevOps build URL

Before analysis, read .\.github\skills\shared\azure-vcpkg-ci-notes.md.

Phase 1: Analyze failed step logs

  1. Inventory evidence — Confirm the build metadata, failed step-log paths, artifact paths, and triplets available from the acquisition step.

  2. Scan step logs — Extract every REGRESSION: line and retain exact failure type keywords: BUILD_FAILED, FILE_CONFLICTS, POST_BUILD_CHECKS_FAILED, and CASCADED_DUE_TO_MISSING_DEPENDENCIES. Capture 2-3 surrounding lines for context.

  3. PR feature-test logs — PR builds may not have REGRESSION: lines. Instead scan for FAIL:, failed with, and feature-test error: lines. Capture verbatim:

    • Compiler errors (missing headers, undefined symbols)
    • Post-build check failures (file path issues, misplaced files)
    • Version validation errors
    • Platform-specific feature guard messages

    Report each feature failure individually. Dependency ports that fail get their own entry.

  4. Version validation — Inspect the "Validate version files" evidence for version database errors.

  5. Write report immediately — Generate and save report.md using the step-log evidence.

Phase 2: Analyze artifacts and enhance

  1. Analyze artifacts — Read relevant stdout-{triplet}.log tails and supporting logs. Classify per references/vcpkg-failure-patterns.md.
  2. Baselines — Check both ci.baseline.txt and ci.feature.baseline.txt.
  3. Enhance report — Add root causes, downstream impact, and actionable recommendations supported by the evidence.

Report Requirements

Format per references/report-template.md:

  • Full build URL: [{buildNumber}](https://dev.azure.com/vcpkg/public/_build/results?buildId={buildId})
  • For PRs: [#{prNumber}](https://github.com/microsoft/vcpkg/pull/{prNumber})
  • List every triplet by full name (e.g., x64-windows, arm64-linux) — never "N triplets". Only include triplets that actually had failures for this specific build.
  • Use exact failure type keywords from logs: BUILD_FAILED, POST_BUILD_CHECKS_FAILED, FILE_CONFLICTS, CASCADED_DUE_TO_MISSING_DEPENDENCIES — never paraphrase
  • Include error messages verbatim — quote the exact text from logs for compiler errors, path issues, and validation failures
  • Dependency ports' failures as separate entries
  • Include baseline/known failures — report them with their failure types, but classify them separately from new regressions

Output Structure

Write reports under the session folder from the environment context, alongside the raw evidence, never into the repository working tree:

{session-folder}/files/ci-failure-analysis/
├── ci-129315/         ← scheduled build
│   └── report.md
└── pr-51202/          ← PR build
    └── report.md

Raw evidence stays where fetch-vcpkg-ci-logs placed it, in {session-folder}/files/ci-failure-logs/; reference those paths from the report instead of copying them.

Critical Rules

  • Do not fetch or download logs as part of this skill; use the acquisition skill first.
  • Scan step logs first because FILE_CONFLICTS may appear only there.
  • Check both baseline files
  • Never suggest <= version constraints or VCPKG_BUILD_TYPE release
  • Clearly distinguish proven root causes from hypotheses and requests for additional diagnostics.

来自 microsoft 的更多技能

oss-growth
microsoft
OSS增长黑客角色
agent-framework-azure-ai-py
microsoft
使用Microsoft Agent Framework Python SDK(agent-framework-azure-ai)构建Azure AI Foundry代理。在创建使用AzureAIAgentsProvider的持久化代理、使用托管工具(代码解释器、文件搜索、网络搜索)、集成MCP服务器、管理对话线程或实现流式响应时使用。涵盖函数工具、结构化输出和多工具代理。
development
airunway-aks-setup
microsoft
在AKS上设置AI Runway——从裸集群到运行模型。涵盖集群验证、控制器安装、GPU评估、提供商设置和首次部署。适用场景:“设置AI Runway”、“接入AKS集群”、“安装AI Runway”、“airunway设置”、“将模型部署到AKS”、“在AKS上进行GPU推理”、“在AKS上配置KAITO”、“在AKS上运行LLM”、“在AKS上使用vLLM”、“在AKS上设置模型服务”、“AI Runway控制器”。
devops
appinsights-instrumentation
microsoft
使用Azure Application Insights对Web应用进行插桩的指南。提供遥测模式、SDK设置和配置参考。适用场景:如何对应用进行插桩、App Insights SDK、遥测模式、什么是App Insights、Application Insights指南、插桩示例、APM最佳实践。
devops
applicationinsights-web-ts
microsoft
使用Application Insights JavaScript SDK(@microsoft/applicationinsights-web)为浏览器/Web应用添加检测。用于真实用户监控(RUM)——页面视图、点击、AJAX/fetch依赖项、异常、自定义事件,以及与后端OpenTelemetry追踪关联的浏览器端GenAI代理追踪。涵盖SDK加载器脚本和npm设置、框架扩展(React、React Native、Angular)、点击分析、遥测初始化器,以及从浏览器发出的代理/工具/模型跨度所遵循的OTel GenAI语义约定。
devops
azure-ai-anomalydetector-java
microsoft
使用适用于 Java 的 Azure AI 异常检测器 SDK 构建异常检测应用程序。在实现单变量/多变量异常检测、时间序列分析或 AI 驱动的监控时使用。
development
azure-ai-language-conversations-py
microsoft
使用azure-ai-language-conversations Python SDK实现对话语言理解(CLU)。当使用ConversationAnalysisClient分析对话意图和实体、构建NLP功能或将语言理解集成到应用程序中时使用。
development
azure-ai-ml-py
microsoft
Azure Machine Learning SDK v2 for Python。用于机器学习工作区、作业、模型、数据集、计算资源和管道。 触发词:“azure-ai-ml”、“MLClient”、“工作区”、“模型注册表”、“训练作业”、“数据集”。
development