New guide
The complete AI operating model for software teams — from first idea to production. Three peer-supervised loops (discovery → build → release) over a catalogue of curated packs: skills, subagents, and hooks, each installed in one line. It's npm for your coding agent. Any agent, any stack — Claude Code, Codex, Cursor, Copilot, Gemini, Kiro.
npx -y skills add eugenelim/agent-ready-repo --skill new-guideAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 14 stars14 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Create or substantially revise user guides, pack pages, and journey pages using Diátaxis plus conversation-first UX. Use when asked to write, simplify, restructure, audit, or modernize tutorials, how-to guides, reference pages, explanations, pack pages, or journey pages so readers can start from a natural-language goal, see what to say, understand what happens next, and reach an outcome without learning internal skill names first. Do NOT use for feature contracts (use `new-spec`), cross-cutting proposals (use `new-rfc`), recording decisions (use `new-adr`), minor single-line edits (normal PR), contributor docs, docstrings, release notes, or blog posts.
SKILL.md
9.4 KB, as published. Nobody here has run it
Guide authoring
Diátaxis determines where information lives. User intent determines how readers enter it.
A reader who does not know any pack or skill names must still be able to begin a real task from the first screen.
Create or substantially revise one user-facing documentation surface. Use Diátaxis to determine the page's job. Use conversation-first design to determine what the reader encounters first.
Prerequisites
- Read the actual skills, commands, or workflows the page documents.
- Identify one primary reader and one primary job.
- Preserve verified commands, permissions, defaults, and side effects.
- For generated pages, edit their source data or template — not the rendered output.
Procedure
Step 1 — Choose mode
Create a new guide, or revise an existing one.
Revise mode applies to substantial changes: restructuring the page's flow, correcting the quadrant, major rewrites, adding conversation-first structure, or auditing for inventory-first writing.
Minor edits — a typo fix, a single updated command, a broken link, a version number — are a normal PR against the existing file. Stop here and redirect to that path; do not proceed with this skill for minor edits.
Step 2 — Pick the slug or confirm the file
- Create: choose a kebab-case slug matching what the reader would have searched
for.
rotate-credential-token, nothow-to-rotate-your-token-step-by-step. The quadrant subdir carries the "how-to" / "tutorial" framing; do not repeat it in the filename. - Revise: confirm the path to the existing file before proceeding.
Step 3 — Write the conversation contract (gated)
Before any prose is drafted, produce the contract and wait for human confirmation. This is the skill's gate — the body is blocked until the contract is signed off.
CONVERSATION CONTRACT:
reader: <role and context — not "all users"; who is reading right now>
job: <the specific thing they are trying to accomplish>
natural_start: <the exact words they would use to begin>
minimum_scope:
- <the minimum the agent needs to start — team, board, time horizon, etc.>
first_result: <the concrete thing the reader gets back>
write_boundary: <what the agent reads vs. what it may change>
next_request: <the most likely follow-up after the first result lands>
Wait for human confirmation or revision. Two outcomes:
- Contract accepted → proceed to Step 4.
- Contract splits into two readers or two jobs → two pages. Surface this and ask the user to pick the first; the second goes to a follow-up.
Step 4 — Choose the page contract type
Determine which of the six surface types applies. For the four Diátaxis types, choose by reader posture — what the reader is doing right now, not what topic they are reading about.
| Reader's posture right now | Surface type |
|---|---|
| On rails, attentive, wants a guaranteed working result | Tutorial |
| Has a named problem, wants the recipe | How-to |
| In a hurry, scanning for the authoritative answer | Reference |
| Away from the keyboard, wants to understand why | Explanation |
Pack pages describe what a pack does and how to start — choose by context. Journey pages walk a complete user flow from first request to final outcome — choose by context.
Step 5 — Load the relevant contract from references/page-contracts.md
Read only the section that matches the surface type chosen in Step 4. Apply its required content and "move lower" rules throughout drafting.
Step 6 — Scaffold (create mode, four Diátaxis types only)
For create mode and the four Diátaxis quadrants, scaffold from the matching asset template and write to the standard destination:
| Surface type | Asset to copy | Destination |
|---|---|---|
| Tutorial | assets/tutorials.md | docs/guides/tutorials/<slug>.md |
| How-to | assets/how-to.md | docs/guides/how-to/<slug>.md |
| Reference | assets/reference.md | docs/guides/reference/<slug>.md |
| Explanation | assets/explanation.md | docs/guides/explanation/<slug>.md |
(Paths are skill-relative — the assets/ folder lives next to this SKILL.md.)
For pack pages and journey pages in create mode, there is no fixed destination; write where the repo's existing guides already live. Match the surrounding structure.
If the repo organizes guides by pack — a docs/guides/<pack>/<quadrant>/ layout
— prefix the destination with the owning pack, or _shared/ for a cross-cutting
guide.
Step 7 — Draft using conversation-first structure
Put the first useful example before inventories, architecture, terminology, or exhaustive options. A reader who reaches word 120 without seeing a prompt, command, or concrete result has waited too long.
Structure the core task flow using:
- Say this — the exact words the reader uses
- What happens — what the agent does in response
- You get — the concrete result
- What to ask next — the likely follow-up
Step 8 — Apply conversation-first rules
Load references/conversation-first.md and
apply its eight sequencing rules. Key checks:
- Realistic user request within the first 120 words
- No more than two product-specific terms before that request
- User language before implementation names
- Read/write boundary explicit
- Next request shown
Step 9 — Edit for density
Load references/clear-prose.md and edit. Pay
particular attention to the ## Conversation-first structure section at the end
— it covers page-level structural tells that a word-level pass will miss.
When your environment provides subagents, hand the draft and the checklist to a read-only subagent; that keeps the style read off your main context. The cold self-read is the floor without subagents.
Step 10 — Run the usability review
Load references/usability-review.md and run
the six-item checklist. Each "yes" is a finding to fix before declaring the draft
ready. Also verify the conversation contract is reflected in the draft.
Step 11 — Check links and cross-link siblings
Apply the link-out rule as you finalize:
- Tempted to explain why mid-tutorial → link to an explanation page
- Tempted to list every option mid-how-to → link to the reference
- Tempted to walk a beginner through setup mid-explanation → link to the tutorial
- Tempted to recommend a best practice mid-reference → link to the explanation
The link can be a placeholder (<!-- TODO: link to … -->) if the adjacent piece
does not exist yet; surface those placeholders in the final summary.
From the new or revised file, add a See also section linking existing siblings
only. From each existing sibling, add a reverse link. Check whether each file
exists before linking — do not write broken links and do not synthesize
plausible-sounding paths.
Do not touch the per-quadrant README.md files — those are the framework's
per-quadrant explainer, not a piece index.
Verification
Before announcing the draft, confirm:
- The first actionable example appears within the first 120 words.
- The reader does not need to know a skill name to begin.
- Read and write behavior is explicit.
- The page shows at least one realistic follow-up.
- Skill inventory appears after user outcomes.
- A complete task has a visible start and finish.
Anti-patterns to refuse
- Drafting before the conversation contract is confirmed. The body is gated. The contract is the cheapest place to catch a mismatched scope; once draft prose is on the page, rework compounds.
- Using this skill for minor edits. A typo fix, a single updated command, a broken link — these are normal PRs. This skill's procedure adds ceremony that minor edits do not warrant.
- Picking the surface type by topic instead of by reader posture. "Authentication" is a topic; learning it, configuring it, looking up its parameters, and understanding why it is cookie-based are four different pieces.
- Blending surface types because "the reader will appreciate the context." They will not. The link-out rule is non-negotiable.
- Drafting a tutorial without running the steps end-to-end. A tutorial that does not produce the promised result is worse than no tutorial. If the steps cannot be run, you are writing a how-to or an explanation — re-open the contract.
- Writing reference in narrative voice. Reference says what; recommendations live in explanation.
- Writing explanation without an About <X> frame. Open-ended explanation absorbs adjacent material and sprawls. Name the question the page answers.
- Leading with a skill or pack inventory. The reader came with a goal. The inventory belongs after the first task completes.