Git搜索
- 作者仓库星标 176,634
- 作者更新于 实时读取
- 作者仓库 flutter
- 领域
- AI 智能
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 88 / 100 · 社区维护
- 作者 / 版本 / 许可
- @flutter · 未声明 license
- Token 消耗评级
- 低消耗
- 接入复杂程度
- 需简单配置
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- macOS · Linux · Windows
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- Shell 执行
- 允许写入 / 修改
- 网络行为
- 允许外网请求
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生,可能与作者实际意图存在差异。
需要注意: 未限定 allowed-tools,默认拥有全部工具权限。
---
name: flutter-pr-checks-finder
description: Find failing checks on a Flutter PR and locate the corresponding LUCI log URLs. You can use the…
category: AI 智能
runtime: 无特殊运行时
---
# flutter-pr-checks-finder 输出预览
## PART A: 任务判断
- 适用问题:提示词、Agent 工作流、模型评估或自动化推理。
- 输入要求:目标材料、限制条件、期望输出和验收方式。
- 证据边界:围绕“Prerequisites / Workflow / 1. Find Failing Checks”读取原文规则,不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断:确认你的需求是否属于提示词、Agent 工作流、模型评估或自动化推理,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“Prerequisites / Workflow / 1. Find Failing Checks”拆成步骤,说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果:给出可复制的命令、文件改动、检查清单或内容草稿,并说明如何继续迭代。
- **04** 风险边界:结合 读取文件、执行终端命令、写入/修改文件、会按任务需要访问外部网络、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、执行终端命令、写入/修改文件;会按任务需要访问外部网络;通常不需要额外 API Key。
- 先小样例验证,再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 原文出现了 `/opt` 这类斜杠命令;如果你的 Agent 支持命令触发,优先用命令开场,再补充目标和边界。
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是:读取文件、执行终端命令、写入/修改文件。
先用一个小任务确认它会围绕“Prerequisites / Workflow / 1. Find Failing Checks”工作;涉及文件或命令时,先看 diff、日志、预览或测试结果。
检查最终产物是否包含明确结果、必要证据和下一步动作;如果输出泛泛而谈,就补充输入、边界和验收标准后重跑。
---
name: flutter-pr-checks-finder
description: Find failing checks on a Flutter PR and locate the corresponding LUCI log URLs. You can use the…
category: AI 智能
source: flutter/flutter
---
# flutter-pr-checks-finder
## 什么时候使用
- 把AI / Agent方向的常用动作沉淀成 Agent 可调用的技能 适合处理AI Agent、提示词、模型评估与自动化推理,核心价值是把输入、判断、执行、验证和交付边界固定下来,避免 Agent 泛泛回答。 把任务拆成可执行、可检查…
- 面向提示词、Agent 工作流、模型评估或自动化推理,优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「Prerequisites / Workflow / 1. Find Failing Checks」组织步骤,不把推断写成作者事实。
- 读取文件、执行终端命令、写入/修改文件;会按任务需要访问外部网络;通常不需要额外 API Key。
- 先跑小样例,确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown,不编造命令、平台或依赖。 作者原文负责流程事实;仓库文件负责来源和命令;流狐只补充适用场景、限制和质量判断。
skill "flutter-pr-checks-finder" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Prerequisites / Workflow / 1. Find Failing Checks
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、执行终端命令、写入/修改文件 | 会按任务需要访问外部网络
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
} Flutter PR Checks Finder
Prerequisites
gh(GitHub CLI) must be installed and authenticated. If not in your PATH, check common locations like/opt/homebrew/bin/ghon macOS orC:\Program Files\GitHub CLI\gh.exeon Windows.- Access to
curlor similar tool to fetch raw logs from LUCI.
Workflow
1. Find Failing Checks
You can use the gh CLI if it's installed and authenticated, or use direct HTTP requests to the GitHub API as a fallback.
Option A: Using gh CLI (Preferred)
Run the following command to list checks:
gh pr checks <PR_NUMBER>
Option B: Using GitHub API via HTTP
If gh is not available, you can use read_url_content or a similar method to interact with the public GitHub API:
- Find the PR SHA:
Make an HTTP request to:
https://api.github.com/repos/flutter/flutter/pulls/<PR_NUMBER>Extract thehead.shafield. - List Check Runs:
Make an HTTP request to:
https://api.github.com/repos/flutter/flutter/commits/<PR_SHA>/check-runsParse the JSON response. CRITICAL: You must handle pagination to avoid missing failures! Check thetotal_countfield. If it is greater than the number of items in thecheck_runsarray (typically capped at 100 or what you set withper_page), make additional HTTP requests by appending?per_page=100&page=<N>to the URL for each subsequent page until all check runs are fetched. Identify all checks that have failed (i.e., whereconclusionisfailure).
Identify all checks that have failed.
2. Retrieve Failure Logs
For each failing check:
Find the Log URL:
- Look for the target URL or link associated with the check.
- The
flutter-dashboardlink typically appears as "View more details on flutter-dashboard" at the bottom of the check view on GitHub. - Alternatively, you can reconstruct the link to the LUCI page based on the name of the failing check and the build number if available.
Example LUCI URL structure:
https://ci.chromium.org/ui/p/flutter/builders/try/<Builder Name>/<Build Number>/overview
Build Raw Log URL:
- Manual: Reconstruct the raw log URL by appending
?format=rawto the log URL or by following the pattern:https://logs.chromium.org/logs/flutter/buildbucket/cr-buildbucket/<Build ID>/+/u/<Step Name>/stdout?format=raw
- Manual: Reconstruct the raw log URL by appending
Fetch Raw Logs:
- Use
curlor similar tool to fetch the content of the raw log URL. - CRITICAL: You must use the raw log URL to avoid HTML formatting and truncated output. Do NOT rely solely on the check summary in the GitHub API, as it may be truncated or lack full context.
- Use
Builder to Step Name Mapping
[!NOTE] Step names can be very specific and hard to guess. This section documents patterns to help find them.
Recipes & Tools
flutter_droneRecipe- Description: Originates from the cocoon or recipes repository. It is typically used for running sharded framework tests and lints (like
analyze,test_general) as specified in .ci.yaml (root). - Pattern:
run test.dart for <shard> shard and subshard <subshard> - URL Transformation: Spaces are replaced by underscores.
- Default: If
subshardis not specified, it defaults toNone. - Example: For
Linux analyze(shard:analyze, no subshard), the URL step name isrun_test.dart_for_analyze_shard_and_subshard_None.
- Description: Originates from the cocoon or recipes repository. It is typically used for running sharded framework tests and lints (like
builder.py& Related Recipes- Description: Found in the
flutter/enginerepository (or recipes repository) and used to execute builds and tests according to target JSON configurations in the builders folder directory. - Pattern: Typically the task name specified in the JSON configuration, often prefixed with
test:. - URL Transformation: Spaces are replaced by underscores.
- Gotcha: If the test name already starts with
test:in the JSON file, the recipe might still add the prefix again (e.g.,test:_test:_Check_formatting).
- Description: Found in the
tester.py- Description: A common helper script in the engine's CI recipes used to execute tests.
- Pattern:
Run <shard> testsorRun <shard> <subshard> tests. - URL Transformation: Spaces are replaced by underscores.
Locating Exact Names in Engine
If guessing fails, find the exact test and task names in the engine configuration:
- Look up the builder in .ci.yaml (engine) to find its
config_nameproperty. - Locate the corresponding JSON file in the builders folder directory.
- Read the JSON file to find the
testsarray and the specifictaskslisted within them.
Fallback
If read_url_content fails with 404 on guessed step names, you may need to find the step name from the LUCI overview page or other sources.
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核