reporting-github-ci-failures
GitHub生成指定 GitHub 仓库的 CI 健康报告，展示当前失败的检查项。支持按分支或 PR 过滤，通过只读方式获取数据并忠实呈现失败详情、链接及修复建议，不执行任何修改操作。
Trigger Scenarios
Install
npx skills add Tracer-Cloud/opensre --skill reporting-github-ci-failures -g -y
SKILL.md
Frontmatter
{
"name": "reporting-github-ci-failures",
"metadata": {
"owner": "Ceren",
"version": "1.1",
"requires": [
"GitHub authentication with read access to the target repository.",
"Explicit repository owner and name for scheduled execution.",
"For recurring delivery, a configured scheduler and destination."
],
"usecases": [
"For maintainers checking which CI checks are failing now in one repository, branch, or PR.",
"For teams receiving recurring CI failure reports at a chosen destination."
],
"last_changed_at": 1789171200,
"last_changed_by": "Jan"
},
"recurring": true,
"description": "Read-only GitHub CI health report of the checks failing right now for one repository, optionally narrowed to a branch or pull request. Not for CI\/CD performance, reliability KPIs, failure rates, or downtime over a period (use analyzing-github-ci-performance)."
}
GitHub CI health
Use this skill to report failing CI checks for exactly one configured GitHub
repository. The schedule must supply owner and repo; it may supply either
branch or pr_number, never both.
The scheduled runner supplies a pre-fetched CI health block. Treat that block as the complete source of truth and return it faithfully as the final report; do not discover a broader organization or repository scope. Preserve every failing check, link, age, responsible PR or branch, coverage notice, and repair handoff in the block.
Tool usage
This skill does not issue GitHub tool calls during unattended execution. Before
the skill runs, the scheduled runner invokes run_github_ci_health, which
collects the configured repository's CI data through read-only GitHub REST GET
requests and supplies the complete rendered report.
Do not perform additional or fallback GitHub discovery. The only interactive
tool used by this skill is propose_scheduled_delivery, which configures the
recurring report and is never called during an unattended run.
This workflow is read-only. Never call fix_github_pr_ci, github_cli,
shell_run, or any other mutating or external-command tool during unattended
execution. Repairs must be requested interactively and explicitly approved.
When offering this report as a recurring task interactively, use kind
recurring_skill and skill name reporting-github-ci-failures. Pass the exact owner and
repo, plus at most one of branch or pr_number, to
propose_scheduled_delivery so confirmation preserves the repository scope.
Interactively, the repository comes from the request; when it names none,
call scan_local_git_workspace and ask with ask_user_choice. Never run
cli_exec (opensre health, opensre integrations …), shell_run, or
github_cli to discover the repository, the token, or the environment: they
print unrelated integration state and are not part of this skill. To show
the checks failing right now during an interactive turn, use
summarize_github_pr_status; this skill's own report is produced by the
scheduled runner only.
Version History
- 4303874 Current 2026-09-22 10:48
