technical-writer-tutorial
暂无简介
- License 许可
- MIT
- 最近更新
- 2026-11-06
- Stars
- ★ 8
- Forks
- 0
02 资源文件 1
- 参考文档evals.json752 B
03 SKILL.md 正文
technical-writer-tutorial
Match the product's actual install/run steps (verify them), don't rewrite adjacent docs, evidence = a reader who follows the tutorial reaches the working end-state without detours.
What you do
Write one tutorial that takes a reader from zero to a working outcome in one sitting. A tutorial is not reference — it is a guided, linear path with one happy path and explicit escape hatches.
- State the outcome up front. The reader should know exactly what they'll have at the end (and roughly how long it takes).
- List prerequisites concretely: OS, versions, accounts, CLIs. "A recent Node" is not a prereq; "Node ≥ 20" is.
- Verify every command yourself or with the owning specialist before shipping. A broken step zero kills the tutorial.
- Each step: one action, the command/code, the expected output. If the output varies, show the anchor line.
- Include a "what could go wrong" per risky step with the recovery command.
- End with "where to go next" — 2–3 concrete follow-up paths.
Output shape
A single markdown file with:
- Outcome — one paragraph on what you'll build + estimated time.
- Prerequisites — versioned list.
- Steps — numbered. Per step: action, code block, expected output, common pitfall.
- Verify — final check the reader runs to confirm success.
- Troubleshooting — 3–6 known failure modes with fixes.
- Next — linked reference docs, manual pages, deeper tutorials.
Store at .guild/runs/<run-id>/docs/tutorial-<slug>.md if tracked.
Anti-patterns
- Skipping setup — "assuming you have the CLI installed" is not a tutorial; it's a trap.
- Assuming tools installed that aren't default (
jq,gh, specific SDK versions). - No expected output — the reader can't tell pass from fail without running six more commands.
- No troubleshooting section — the first error sends them to support.
- Multiple happy paths in one tutorial (mac/linux/windows all fully inline) — fork into variants instead.
- Reference-doc creep — linking out is better than inlining the full schema.
Handoff
Return the tutorial path to the invoking technical-writer specialist. If the tutorial surfaces a broken step or missing feature, raise a followup for the owning engineering specialist. Chain into technical-writer-api-docs for reference depth and technical-writer-user-manual for task-based follow-ups. This skill does not dispatch.
04 同仓库更多 Skills
social-media-thread
暂无简介
security-dependency-audit
暂无简介
marketing-launch-plan
暂无简介
marketing-ab-copy-variants
暂无简介
frontend-bundler-config
暂无简介
doc-writer-onboarding-doc
Authors onboarding documentation and contributor guides — getting-started narrative pages, CONTRIBUTING.md, developer onboarding docs, new-user setup guides, CODE_OF_CONDUCT.md, GOVERNANCE.md. Pulled by the `doc-writer` specialist. TRIGGER: "write a getting-started guide", "create a CONTRIBUTING.md", "onboarding guide for new developers", "write the contributor guide", "new-user setup documentation", "developer onboarding doc", "write the getting started page", "add a CODE_OF_CONDUCT". DO NOT TRIGGER for: project README overview (use `doc-writer-readme`), onboarding email sequences for existing SaaS users (use `copywriter-email-sequences` — copywriter owns lifecycle nurture emails), API reference docs (use `technical-writer-api-docs`), formal step-by-step tutorials with prerequisite/verify blocks (use `technical-writer-tutorial`), changelogs or release notes (use `technical-writer-release-notes`), marketing copy about onboarding (use `copywriter-long-form`).