[cli-tools-test] [CLI exploratory testing] Windows deprecated missing-tool key fails strict dry-run gate

未关闭 适合新手
  • 哪里不对？
#74,755 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
2/5
预计耗时
1-3 小时
新手友好度
70/100
Issue 类型
缺陷
描述清晰度
描述清楚
活跃度
活跃
技术栈
github-actions, go
领域
ci-cd

调研方向

阅读 .github/workflows/windows.md 第54行附近的内容，并检查可信的 workflow 示例如何配置安全输出。将已弃用的键 missing-tool 迁移为 report-missing-tool，然后使用严格的 dry-run 设置编译 Windows workflow，并检查重新生成的 windows.lock.yml。批处理警告门通过且 lock 文件反映更新后的 workflow 时，即为完成。

由索引模型根据 Issue 内容生成。

描述

automation cli-tools testing
Overview

Bounded testing on 2026-11-04 found that .github/workflows/windows.md:54 uses the deprecated missing-tool safe-output key. Its strict dry-run compile emits a lock but fails the overall warning gate. Severity: Low; observed once. This is a repository workflow compatibility issue, not evidence that the safeguard should be weakened.

Reproduction
  1. On github/gh-aw, invoke agenticworkflows MCP compile with workflows=[windows], dry_run=true, strict=true.
  2. Inspect both workflow and batch results.
  3. The workflow result is valid and identifies windows.lock.yml, but the compiler batch is invalid and dry_run.gate is failed.

Expected: the maintained trusted sample uses current settings and passes the strict gate. Actual diagnostic: The missing-tool safe-output key is deprecated. Use report-missing-tool instead.

Proposed correction: migrate the setting through normal workflow maintenance, then compile and review the regenerated lock. No workflow source was edited by this tester.

Diagnostics and bounded coverage

Passed checks

  • Discovery returned 356 workflows, all marked compiled. Sampled triggers and imports matched the selected sources; freshness was not independently established.
  • Five logs queries, always count=3: last day; daily-workflow-security-model over seven days; copilot engine over one day; nonexistent-workflow-xyz; future date 2099-01-01. Exposed JSON summaries were readable. Nonempty results had at most three runs and consistent IDs, workflow names, engines, statuses and timestamps. Unknown workflow returned actionable guidance; future date returned zero runs. Count-limited results warned about partial coverage.
  • Two completed-run audits: failed Daily Workflow Security Model and successful PR Finisher. Metadata agreed with logs. Reports exposed job timing, metrics and safe-output summaries; the success sample exposed bash and safeoutputs.noop calls, MCP health, and 24 network requests. Its noop described unfinished task work, so workflow success is not task completion. Neither sample created GitHub resources. Invalid audit ID 0 returned clear guidance.
  • Windows emitted a local lock with frontmatter/body hashes, strict/dry-run metadata, expected job sections and staged safe outputs; ShellCheck passed with zero added findings. This was structural inspection, not independent complete Actions validation.
  • Logs took approximately 7.5–9.7 seconds for nonempty queries and 1.5 seconds for the future date. Audits took 11.0 and 9.4 seconds. Compiles took 0.9–2.7 seconds. These are single observations.

Observed problems

Windows failed the batch warning gate on the deprecated key. A per-workflow valid result must not be interpreted as an overall gate pass. The tool also warned that cached model inventory may be stale.

Blocked or unfinished required checks

All three targeted compiles used dry_run=true and strict=true; zero passed the overall gate. Archie was rejected for pre-agent steps and custom MCP servers. Daily Workflow Security Model was rejected for custom steps. These safeguard denials were not retried or worked around. Passing complex/import compilation, corresponding generated-lock inspection, and compile-to-existing-run integration remain unverified. The failed existing run used SHA f3dc4385203cb61f4ebc410546ee91822538e884; it does not validate a newly generated lock.

Not tested (expected exclusion)

  • Malformed YAML, invalid tool configuration and missing imports: synthetic fixtures prohibited.
  • Private raw logs, raw directory structure and independent artifact completeness: outside scope.
  • Execution of newly compiled locks: dispatch prohibited.
  • Historical/retention and in-progress audit variants: outside the two completed-run sample.
  • Redundant date variants and incremental recompilation: optional variants omitted after failed compile gates.
  • Created-resource links and failed sub-agent attribution: not exercised by these samples.
Next actions

Migrate the Windows setting and review which trusted samples satisfy the existing safeguards. Required compile coverage is incomplete and is being reported separately. No locks were published, pushed or executed. This sample does not establish repository-wide CLI health.

References:

Generated by 🧪 Daily Cli Tools Tester · codex · gpt60 · ~155 AIC · ⌖ ~57 AIC · ⊞ 21.2K · ◷

  • expires on Nov 10, 2026, 9:23 PM UTC-08:00
主要语言
Go
星标
5.6k
派生
615
平均合并
5 小时 19 分钟
30 天内合并 PR
798

环境准备

在 Codespaces 中打开

在浏览器里用你自己的 GitHub 账号启动这个项目的开发容器。

  • 提供 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 阅读贡献指南

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

github/gh-aw 的其他 Issue

查看 github/gh-aw 的全部 Issue

相似的 Issue

更多 Go Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。