Source profileQuality 86/100

eugenelim/agent-ready-repo/.agents/skills/author-product-docs/SKILL.md

author-product-docs

Create, revise, retrofit, audit, or verify product documentation — pack READMEs, journeys, tutorials, how-to guides, reference pages, and explanations. Use when asked to write, improve, restructure, audit, or verify user-facing documentation, fix a pack README, create a guide for a feature, update a journey page, or check whether docs match shipped behavior. Infers the mode from the request. Do NOT use for feature specifications (use new-spec), cross-cutting proposals (use new-rfc), decisions (u

Source repository stars
15
Declared platforms
0
Static risk flags
2
Last source update
2026-08-05
Source checked
2026-08-05

Decision brief

What it does—and where it fits

Diátaxis determines what a page does for the reader. Canonical behavior determines what it says.

Best for

  • Use when asked to write, improve, restructure, audit, or verify user-facing documentation, fix a pack README, create a guide for a feature, update a journey page, or check whether docs match shipped behavior.

Not for

  • Making product claims without inspecting the canonical source. Read the skill source before writing what it "can do."
  • Writing to docs/guides/ for external product users. docs/guides/ is for repo maintainers. External guides live in guides/.

Compatibility matrix

Platform support, with evidence labels

PlatformStatusEvidenceWhat to check
CodexNot declaredNo explicit evidencePortability before use
Claude CodeNot declaredNo explicit evidencePortability before use
CursorNot declaredNo explicit evidencePortability before use
Gemini CLINot declaredNo explicit evidencePortability before use
Open the compatibility checker

Installation

Inspect first. Install second.

The source command is displayed only when detected. A safe inspection prompt is always available so your agent can explain every action before execution.

Source-detected install commandSource
npx skills add https://github.com/eugenelim/agent-ready-repo --skill ".agents/skills/author-product-docs"
Safe inspection promptEditorial

Inspect the Agent Skill "author-product-docs" from https://github.com/eugenelim/agent-ready-repo/blob/9563bc93aa5b0750b327be2fd95676ff2a5ec63b/.agents/skills/author-product-docs/SKILL.md at commit 9563bc93aa5b0750b327be2fd95676ff2a5ec63b. List every install step, command, network request, credential, file read/write, external action, and rollback step. Explain whether it fits my task. Do not install or execute anything until I approve.

Workflow

What the source asks the agent to do

  1. 01

    Procedure

    Infer the mode from the request. Do not require the user to name it.

    External audience (product users): document in guides// (this catalogue) or the adopter's configured guide root.Internal audience (repo maintainers/contributors): document in docs/guides/ (this catalogue) or the adopter's internal docs location.pack.toml — name, description, version, scope, dependencies, first-value
  2. 02

    Step 1 — Resolve the mode

    Infer the mode from the request. Do not require the user to name it.

    Infer the mode from the request. Do not require the user to name it.When a request is ambiguous between create and revise, read the target file first. If it exists and is substantive, treat as revise. If absent or near-empty, treat as create.
  3. 03

    Step 2 — Resolve the documentation audience

    Before drafting anything, confirm the documentation is for an external catalogue or product user — not internal maintainer guidance. The two ownership trees are distinct:

    External audience (product users): document in guides// (this catalogue) or the adopter's configured guide root.Internal audience (repo maintainers/contributors): document in docs/guides/ (this catalogue) or the adopter's internal docs location.Before drafting anything, confirm the documentation is for an external catalogue or product user — not internal maintainer guidance. The two ownership trees are distinct:
  4. 04

    Step 3 — Resolve the target artifact

    Identify the specific artifact:

    Identify the specific artifact:For retrofit mode, identify the connected set: entry surfaces, related guides, pack README, and journey.When the artifact is ambiguous, record a defensible assumption and continue — do not add a mandatory checkpoint unless uncertainty would materially change audience, behavior, target artifact, a destructive claim, or the…
  5. 05

    Step 4 — Inspect canonical behavior before drafting

    Before writing any product claim, read the authoritative sources:

    pack.toml — name, description, version, scope, dependencies, first-valueActual .apm/skills//SKILL.md — modes, inputs, outputs, read/write behaviorSchemas, permissions, and result limits in the skill source

Permission review

Static risk signals and limitations

Reads files

low · line 30

The documentation asks the agent to read local files, directories, or repositories.

When a request is ambiguous between create and revise, read the target file first. If it exists and is substantive, treat as revise. If absent or near-empty, treat as create.

Writes files

medium · line 30

The documentation asks the agent to create, modify, or delete local files.

When a request is ambiguous between create and revise, read the target file first. If it exists and is substantive, treat as revise. If absent or near-empty, treat as create.

Writes files

medium · line 124

The documentation asks the agent to create, modify, or delete local files.

*For adopter repositories:** inspect existing guide locations first. Ask once if structure is absent and the write destination would determine the artifact's type. Write to the structure the repo already uses; don't impose this catalogue's

Evidence record

Why each signal appears

EvidenceSourceComputedTestedEditorial
SignalValueEvidence typeMeaning
Quality score86/100ComputedDocumentation, specificity, maintenance, and trust rules
Repository stars15SourceRepository attention, not individual Skill quality
Compatibility0 platformsSourceDeclared in the catalog source record
Usage guideautomated source guideEditorialGenerated or reviewed according to the visible evidence level

Pinned source

Provenance and original SKILL.md

Repository
eugenelim/agent-ready-repo
Skill path
.agents/skills/author-product-docs/SKILL.md
Commit
9563bc93aa5b0750b327be2fd95676ff2a5ec63b
License
Apache-2.0
Collected
2026-08-05
Default branch
main
View the original SKILL.md

Product documentation authoring

Diátaxis determines what a page does for the reader. Canonical behavior determines what it says.

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 work with product documentation — pack READMEs, journeys, and Diátaxis guides — grounded in what the product actually ships today.

Output rendering

Rationale / narrative — Use short ## headings and 2–3 sentence paragraphs. Don't force narrative into a table. Key–value / one record — For a single record's fields, use an aligned key: value list, not a two-row table. Status list — Lead each row with a status glyph (● running, ✓ done, ○ idle, ⚠ blocked).

Procedure

Step 1 — Resolve the mode

Infer the mode from the request. Do not require the user to name it.

ModeSignals
Create"write a guide", "new tutorial", "create a README", "document this feature"
Revise"improve", "update", "rewrite", "restructure", "fix", "simplify"
Retrofit"connect these pages", "fix the journey", "reorganize the docs", "make it coherent"
Audit"audit", "review", "what's missing", "what's wrong", "check quality"
Verify"does this match what ships", "check accuracy", "verify against behavior"

When a request is ambiguous between create and revise, read the target file first. If it exists and is substantive, treat as revise. If absent or near-empty, treat as create.

Step 2 — Resolve the documentation audience

Before drafting anything, confirm the documentation is for an external catalogue or product user — not internal maintainer guidance. The two ownership trees are distinct:

  • External audience (product users): document in guides/<pack>/ (this catalogue) or the adopter's configured guide root.
  • Internal audience (repo maintainers/contributors): document in docs/guides/ (this catalogue) or the adopter's internal docs location.

If the request describes a maintainer workflow (CI debugging, seed authoring, adapter maintenance, internal tooling), it belongs in docs/guides/ — not guides/. See references/repository-ownership.md.

Step 3 — Resolve the target artifact

Identify the specific artifact:

ArtifactUse when
Pack READMEPrimary landing and discovery doc for a pack
JourneyComplete user flow from first request to final outcome
TutorialBeginner needs a guaranteed working result from scratch
How-to guideCompetent reader has a specific named problem to solve
ReferenceReader needs authoritative, dry, complete fact lookup
ExplanationReader wants to understand why something works the way it does
Guide index / landingEntry surface linking into related guides

For retrofit mode, identify the connected set: entry surfaces, related guides, pack README, and journey.

When the artifact is ambiguous, record a defensible assumption and continue — do not add a mandatory checkpoint unless uncertainty would materially change audience, behavior, target artifact, a destructive claim, or the canonical source.

Step 4 — Inspect canonical behavior before drafting

Before writing any product claim, read the authoritative sources:

  • pack.toml — name, description, version, scope, dependencies, first-value
  • Actual .apm/skills/<name>/SKILL.md — modes, inputs, outputs, read/write behavior
  • Schemas, permissions, and result limits in the skill source
  • README.md (current) — what exists already
  • Journey files (JOURNEY.md if present)
  • Related user guides
  • DESIGN.md if present — for verified architecture claims only

Do not make product claims about what a skill "can do" without reading its source. A claim that survives without this inspection is not a product claim — it is a hallucination.

Step 5 — Write the documentation contract

Before drafting, write a short internal contract. This is not a mandatory user checkpoint — record it as a comment block in your reasoning, not as a human-confirmation gate (unless uncertainty about audience or behavior is blocking you).

mode: <create | revise | retrofit | audit | verify>
audience: <external product user | internal maintainer>
situation: <what the reader is in the middle of>
primary job: <the specific thing they are trying to accomplish>
natural start: <the exact natural-language request they would use>
expected result: <the concrete thing they get back>
human decision: <what remains theirs to decide>
read/write boundary: <what the skill reads vs. what it may change>
canonical sources inspected: <list the files you read>
page kind: <pack README | journey | tutorial | how-to | reference | explanation | index>
journey association: <what journey this page belongs to, if any>
likely next: <the most likely next request after this artifact>

Step 6 — Assign the page kind via the Diátaxis compass

For guide artifacts, assign one kind from reader posture — what the reader is doing right now, not what topic they are reading about:

Reader's posture right nowKind
On rails, attentive, wants a guaranteed working resulttutorial
Has a named problem, wants the recipehow-to
In a hurry, scanning for the authoritative answerreference
Away from the keyboard, wants to understand whyexplanation

This is a page contract, not a directory choice. Load the matching contract from references/page-contracts.md and apply it throughout drafting.

Step 7 — Select the minimum useful artifact set

Default to ONE artifact. Do not:

  • Create sibling pages merely to fill the other Diátaxis kinds
  • Create empty category directories
  • Update a README, index, or journey unless the new work materially changes discovery or the canonical flow

A single well-executed how-to is more useful than four thin quadrant stubs.

Step 8 — Resolve the write destination

Determine where to write the artifact. This skill is portable — it must not hardcode this catalogue's specific paths.

For this catalogue (agent-ready-repo):

  • External product guides: guides/<pack>/<kind>/<slug>.md
  • Pack README: packs/<pack>/README.md
  • Journey: packs/<pack>/JOURNEY.md (if convention is established)
  • Internal maintainer guides: docs/guides/<kind>/<slug>.md

For adopter repositories: inspect existing guide locations first. Ask once if structure is absent and the write destination would determine the artifact's type. Write to the structure the repo already uses; don't impose this catalogue's layout.

See references/repository-ownership.md for the full ownership model.

Step 9 — Draft task-first

Structure the core task flow for user-facing documentation:

  • What the user can accomplish — the goal, in the user's own language
  • What to say or do — the natural-language request or action
  • What the system reads or changes — the read/write boundary
  • What result the user receives — concrete, verifiable
  • What decision remains theirs — human in the loop
  • What to do next — the likely follow-up

Put a realistic user request within the first 120 words. No more than two product-specific terms before it.

Load references/conversation-first.md and apply its eight sequencing rules.

Step 10 — Format reference material compactly

For reference pages, keep lookup material structured and scannable: aligned key-value lists for single records, tables for sets of comparable items. Apply the contracts from references/page-contracts.md.

Step 11 — Edit for density

Load references/clear-prose.md and edit. Cut hedges, uniform rhythm, throat-clearing openers, inflated verbs. Check structural tells: treadmill effect, symmetrical padding, false precision.

Step 12 — Cross-link only existing artifacts

Link to existing files or files created in the same change. Verify file existence before writing a link. Surface missing sibling links as <!-- TODO: link to … --> rather than writing broken links.

For pack READMEs: link to the pack's guide home. For guides: link to related siblings that exist. For journeys: link to the pack README and relevant how-to guides.

Step 13 — Render and verify

When a renderer is available, build the documentation and verify the output before reporting done. Apply proportionate verification from references/rendered-verification.md:

  • Content-only edits: link check only
  • Navigation changes: route check
  • Page-layout changes: visual review of rendered output

For audit mode: produce evidence-based findings without editing the source. List specific files, lines, and what was found. Do not edit unless implementation was explicitly requested alongside the audit.

For verify mode: read canonical behavior sources, then check each documentation claim against them. List verified claims, unverified claims, and claims that contradict current behavior.

Step 14 — Report

At the end, report:

  • Mode used and why it was inferred
  • Artifact decision (kind, slug, destination)
  • Canonical sources inspected
  • Files changed
  • Verification performed
  • Unverified behavior (claims you could not confirm)
  • Deliberately omitted artifacts

Anti-patterns to refuse

  • Making product claims without inspecting the canonical source. Read the skill source before writing what it "can do."
  • Writing to docs/guides/ for external product users. docs/guides/ is for repo maintainers. External guides live in guides/.
  • Imposing guides/tutorials/, guides/how-to/ etc. in an adopter repo that doesn't use that structure. Inspect first; match what exists.
  • Creating four Diátaxis pages when one was asked for. Select the minimum useful artifact. One complete page beats four thin stubs.
  • Creating empty category directories. Write the artifact, not the container.
  • Picking the Diátaxis kind by topic instead of reader posture. "Authentication" is a topic. Whether the reader is on rails (tutorial), has a problem (how-to), needs a fact (reference), or wants to understand (explanation) determines the kind.
  • Drafting before knowing the audience. Internal maintainer guidance written to guides/ ends up shipped to adopters.
  • Editing rendered output. The source is the canonical artifact. Edits to web/ or docs-site/ generated output don't survive the next build.
  • Claiming rendered verification without running the renderer. Only report verification that actually ran.

Alternatives

Compare before choosing

Computed 9823,835

alirezarezvani/claude-skills

quality-manager-qms-iso13485

ISO 13485 Quality Management System implementation and maintenance for medical device organizations. Provides QMS design, documentation control, internal auditing, CAPA management, and certification support. Use when working with medical device quality systems, preparing for ISO 13485 audits, managing regulatory compliance documentation, setting up corrective actions, or building audit preparation programs. Useful for quality management, audit preparation, regulatory compliance, medical device d

Computed 9810,896

huggingface/skills

huggingface-zerogpu

AI demos and GPU compute with Gradio Spaces and Hugging Face Spaces ZeroGPU. Use when writing or reviewing code that uses `@spaces.GPU`, configuring `python_version` or `requirements.txt` for a ZeroGPU Space, or handling ZeroGPU-specific code constraints — pickle-based process isolation, `gr.State` semantics across the worker boundary, no `torch.compile` (use AoTI instead), CUDA wheel-only builds (no `nvcc` at build or runtime), large vs xlarge sizing, and dynamic duration callables. Make sure t

Computed 9732,671

K-Dense-AI/scientific-agent-skills

biopython

Comprehensive molecular biology toolkit. Use for sequence manipulation, file parsing (FASTA/GenBank/PDB), phylogenetics, and programmatic NCBI/PubMed access (Bio.Entrez). Best for batch processing, custom bioinformatics pipelines, BLAST automation. For quick lookups use gget; for multi-service integration use bioservices.

Computed 9732,671

K-Dense-AI/scientific-agent-skills

esm

Use when working directly with the `esm` Python SDK, ESM3 or ESMC model IDs, Forge/Biohub inference clients, or ESMFold2 folding workflows.