Docs & Knowledge · nexu-io/html-anything
eng-runbook
Service overview + alerts table + dashboards + operational commands + on-call + incident checklist
L1 Basic · Source registration onlyV1 Commit-PinnedDual-pass Audit
Control levels describe skill capabilities and suggested control measures; they do not represent quality or safety certifications, nor do they dictate usability. Usage decisions rest entirely with users and administrators. Different agent hosts apply varying constraints; do not assume hosts provide automatic protection. Third-party skill copyrights belong to their original authors. Use at your own risk.
1How It Works & Mechanism
This skill is not an executable program, but rather follows the concept of 'one directory = one template': SKILL.md's frontmatter is used for the selector display, and its body is the prompt sent to the model; adding a template means adding a directory, requiring no changes to TypeScript.
next/src/lib/templates/loader.ts
routes rescan disk and the client refetches `/api/templates`.
The header comments of the same file specify the directory convention: SKILL.md (frontmatter + prompt body) / example.md (optional input sample) / example.html (optional preview). This skill directory strictly follows this structure.
The actual prompt sent = platform SHARED_DESIGN_DIRECTIVES + SKILL.md body of this skill + user content, assembled into a single prompt via assemblePrompt.
next/src/lib/templates/shared.ts
Wrap a per-template instruction body with the shared design directives and
Therefore, the body of this skill is only responsible for the 'layout/style/component library' section; output discipline (single-file HTML, mandatory use of real user data, content-driven quantity) comes from the shared directives, eliminating the need for template authors to declare them repeatedly.
Final drafts bypass the file system: shared directives explicitly forbid file tools like Write/Edit/Bash, requiring the model to stream the entire HTML directly as the response body (stdout is captured and written to disk by the front end).
next/src/lib/templates/shared.ts
**禁止使用 Write / Edit / MultiEdit / Bash / Create / 任何文件系统工具**。不要把 HTML 写到任何 \`.html\` 文件里。前端直接捕获你的 stdout 文本, 文件落盘由前端负责。
This is the source of the 'zero disk writes by the skill itself' mechanism: the host front end writes the files, not the skill; there are no scripts inside this directory.
The page skeleton follows 'numbered sections + tables + command blocks': the five sample sections in order are Service summary, Alerts, Common procedures, On-call rotation, and Incident response.
next/src/lib/templates/skills/eng-runbook/example.html
<h2><span class="index">03</span>Common procedures</h2>
Each section title uses a small monospace .index number (01-05) to maintain visual rhythm, corresponding to the first five items in the SKILL.md layout list; the sample topic is a fictional auth-service.
The alert table is the core component of this template: four columns (Alert / Severity / What it means / First response), providing step-by-step instructions on 'what to do first upon being paged'.
next/src/lib/templates/skills/eng-runbook/example.html
<thead><tr><th>Alert</th><th>Severity</th><th>What it means</th><th>First response</th></tr></thead>
Inconsistent with the 'severity / threshold / runbook link' three-column layout in SKILL.md: the sample omits threshold and runbook link, replacing them with What it means + First response (more action-oriented, see second_pass).
Command blocks are organized by 'operation name + when to use + explanation + copyable command', where the sample uses a leading $ and comment span to mark command lines ready for direct copying.
next/src/lib/templates/skills/eng-runbook/example.html
<span class="cmt"># Deploy auth-service v4.7.3 to production</span>
Corresponds to SKILL.md's 'Common procedures (mono code block, one-click copy)'—however, the sample lacks a copy button (script-free), so 'one-click copy' relies on the host's code block copy capability or manual selection.
The on-call rotation schedule is laid out by month with four columns (Week / Primary / Secondary / Backup (escalation)), featuring one row per week across four weeks.
next/src/lib/templates/skills/eng-runbook/example.html
<thead><tr><th>Week</th><th>Primary</th><th>Secondary</th><th>Backup (escalation)</th></tr></thead>
SKILL.md specifies 'On-call rotation (this week + next week)', whereas the sample provides the full four weeks of the month + a three-tier escalation chain, offering richer information and different granularity (see second_pass).
The visuals adopt a dark on-call console style: dark blue-black background #0c0e14, panels #14171f, mint green accent #6ee7b7, supplemented with warning yellow and danger red.
next/src/lib/templates/skills/eng-runbook/example.html
--bg: #0c0e14;
A dark background combined with high-contrast status colors is ideal for on-call scenarios (nighttime viewing, auxiliary screen next to a terminal); SKILL.md does not enforce a theme, which is set by the sample and represents an area of greater template freedom.
Incident response presents action items for the 'first 30 minutes' in a two-column checklist grid, while the dependency list uses ok/warn two-state coloring to convey dependency health.
next/src/lib/templates/skills/eng-runbook/example.html
.checklist { display: grid; grid-template-columns: 1fr 1fr; gap: 14px; }
Corresponds to SKILL.md's 'Incident response checklist' and 'Service overview (topology + dependencies)'; dependency rows use .deps li to hold module names + ok/warn indicators.
The server-side /api/convert performs only three tasks: read the skill from disk by templateId, assemble the prompt, spawn the local agent CLI, and stream the output via SSE forwarding.
next/src/app/api/convert/route.ts
prompt = assemblePrompt({ body: skill.body, content, format });
In the same file, it first executes `const skill = loadSkill(templateId);` (returning 400 if the template does not exist) before passing it to invokeAgent; the skill directory itself spawns no processes and performs no networking.
2Core Capabilities
01Actionable on-call runbook single-page: overview / alerts / procedures / on-call rotation / incident checklist
02Alert-to-action lookup table (Alert/Severity/What it means/First response)
03Service overview panel: text summary + dependency list (including ok/warn status indicators)
04Procedure blocks for common operations (including 'when to use' hints and copy-ready shell commands)
05Full-month on-call schedule + three-tier escalation chain (Primary/Secondary/Backup)
06Incident response checklist (first 30 minutes, two-column grid)
4Recommended Governance & Evidence
Suggested Controls for Safe Operation
For Agent Runtime
Can be auto-loaded
For Administrators
Register source and version
Evidence & Triggers (File Paths & Excerpts)
- SKILL.md: Pure prompt directives and rule text, without script execution or external impact
Audit Notes: Grading target = the skill directory itself (SKILL.md + example.html/example.md; this directory contains zero scripts, zero credential access, zero disk-write directives, and no network calling code). According to the Batch 4 final criteria, html-anything templates are uniformly graded green: SKILL.md does not reference any runtime service backend/API across its entire text, nor does it send user materials to external generation backends (no orange criteria); no reverse engineering, impersonation, or bypassing of backends (no red criteria). The only endpoints appearing in the directory are external links in artifact examples (none (example.html contains zero external links across its entirety: font family relies on the system stack Inter/ui-monospace, with no font links, no CDNs, and no image URLs)), which are classified as 'artifact examples' and are not subject to grading per criteria, listed here merely for self-attestation. The host side should also note: preview iframes use sandbox="allow-scripts allow-same-origin", the local agent CLI is spawned (next/src/lib/agents/invoke.ts), and shared directives require the final output to reference CDNs—these are platform behaviors mentioned solely for clarification.
5Second-Pass Verification
- [unlocatable] 「Dashboards links 卡片」 — 样例五节中无 dashboards/Grafana 类链接卡片，全文 grep dashboard 零命中；SKILL.md 承诺的这一节在样例里无对应实现（既非改名也非隐藏，属未实现）。
- [discrepancy] 「Alerts table (severity / threshold / runbook link)」 — 表列实为 Alert / Severity / What it means / First response，没有 threshold 列与 runbook 链接列。
- [discrepancy] 「Common procedures (mono 代码块, 一键复制)」 — mono 代码块成立（--mono + 前置 $ 与 cmt span），但「一键复制」需 JS，样例零 <script>，无复制按钮。
- [discrepancy] 「On-call rotation (本周 + 下周)」 — 样例给整月四周（Oct 27–Nov 23）且含 Backup (escalation) 列，比承诺更全但粒度不同。
- [ok] 「Incident response checklist」 — 第 05 节 .checklist 双列网格承载「最初 30 分钟」动作项。
- [ok] 目录内零脚本/零端点 — 无 <script>、无外链字体与图片、无 CDN；命令均为虚构内部工具示意。
- [ok] 深色主题 — --bg:#0c0e14 / --paper:#14171f / --accent:#6ee7b7 / --warn:#fbbf24 / --danger:#f87171（SKILL.md 未规定主题，属模板自由度）。
6Verdict & Scenarios
Best for: Ideal for SRE/backend teams creating on-call runbooks for single services; suitable for condensing alert documentation scattered across Confluence into a single printable/projectable runbook; ideal for rapidly generating 'current state + first response' pages prior to incident post-mortems.
Not for: Not suitable for multi-service overviews (this template is a single-service long page); not suitable for external publication or customer-facing documentation (dark theme and internal commands should not be shared externally); not suitable for scenarios requiring true interactivity (copy buttons, linked charts); also not suitable for non-engineering readers (high density of terminology and commands).
Install Agent-Ready
① Verified Mirror 2026-09-15
Option A · Direct DownloadDownload eng-runbook.tar.gz
sha256:
Option B · Copy Guide for Agentb08657d6c60cbf74…Agents automatically download and verify mirrors from this JSON guide.
② Upstream GitHub
GitHub Repo ↗
Pinned to commit 553ed98c28.