article
Skill4 starsChanged 6 months agoMITView on GitHub
Generate a publication-ready article for any site, in any language. Composes the anthropic-skills pipeline (content-brief → write-content → eeat-audit) and applies your site-config.yaml — brand voice, audience, terminology, language profile — as drafting constraints. Use when the operator says "write an article about X", "/article <keyword>", or wants to publish a new blog post. Default language is English; switchable via site-config.yaml.
10 other projects use a version of this.
Adopted by
Most recently taken up by
- parker-brown-family/agent-skillstheir own version, first seen Nov 8, 2026
- RightBlogger/bloggingskillstheir own version, first seen Nov 1, 2026
- aaronjmars/miroshark-aeontheir own version, first seen Oct 21, 2026
- aeonfun/aeontheir own version, first seen Oct 20, 2026
- MiroShark/miroshark-aeontheir own version, first seen Oct 17, 2026
- durchnull/rendertheir own version, first seen Oct 15, 2026
Used with
Projects with it also use aeon-update, auto-workflow and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
aaronjmarsSkillaeon-update
Pull framework updates from the upstream Aeon repo into this instance - 3-way merges canon's new commits into a PR, never clobbering operator config.
Also in 4 projects that use this
aaronjmarsSkillauto-workflow
Two-mode aeon.yml workflow builder - analyze inspects URLs and emits a tiered, signal-verified skill-enablement plan plus an aeon.yml diff; enable flips slugs to enabled:true and opens a PR.
Also in 4 projects that use this
- +10 more from aaronjmars
Other versions10 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 10
- aeonfun/aeon★ 793same name
- aaronjmars/aeon★ 789same name
- aaronjmars/miroshark-aeon★ 20same name
- MiroShark/miroshark-aeon★ 20same name
- ariadoss/superskills★ 12same name
- RightBlogger/bloggingskills★ 4same name
- petestewart/claude-skills★ 2same name
- r4yb3l/tokendial★ 2same name
- parker-brown-family/agent-skills★ 1same name
- durchnull/render★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repositoryOne other file it gives its agents.
- language-reviewSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. 0 are common, 8 are distinctive, and 6 are very specific to this project.
0
Common
Most of the 9 edited versions we read keep these.
8
Distinctive
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 25
If it isn't installed yet, tell the operator to run that command first, then re-invoke this skill.
- Distinctive: line 30
Before starting, READ these files (they are the brand context and MUST be respected):
- Very specific: line 43
If site-config.yaml is missing, STOP and tell the operator to copy examples/site-config.example.yaml from this repo into their project root.
- Distinctive: line 62
Run these steps in order. Each step uses the named upstream skill.
- Distinctive: line 101
Never invent prices. Never reference legacy prices.
- Distinctive: line 105
Always quote the full sequence when relevant.
- Very specific: line 111
Use site.positioning once, naturally, in the article.
- Very specific: line 150
If site-config.yaml has regulatory_topic: true (e.g.
- Distinctive: line 164
If a claim cannot be verified against a citable source, EITHER cite the source or remove the claim.
- Very specific: line 183
Write the final article to <output.path>/<slug>.md with frontmatter matching output.frontmatter_format.
- Distinctive: line 244
If the keyword has no clear search intent for the configured audience (e.g.
- Distinctive: line 246
Don't write thin content to fill quota — better to write 0 articles than 1 that erodes brand.
- Very specific: line 249
If write-content returns content that violates hard rules (terminology, banned vocabulary, language profile, phasing dates) and a single …
- Very specific: line 254
If site-config.yaml is incomplete (missing voice.tone, missing audience.primary, etc.), STOP and ask the operator to fill it in before …
“Common” is measured: kept by at least 6 in 10 of the 9 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
article — generalized SEO article generator
This skill composes the upstream content-brief, write-content, and
eeat-audit skills with site-specific brand context loaded from
site-config.yaml, then emits a markdown file ready to drop into your
content directory.
It is the generalized form of FakturaFlow's fakturaflow-article skill.
The pipeline shape is the value; the FakturaFlow-specific bits are
config.
Prerequisites
The upstream anthropic-skills plugin must be installed:
/plugin install anthropic-skillsIf it isn't installed yet, tell the operator to run that command first, then re-invoke this skill.
Required pre-checks
Before starting, READ these files (they are the brand context and MUST be respected):
site-config.yamlat project root — brand voice, audience, terminology, output path, language profile selection..claude/skills/article/language-profiles/<profile>.md— the active language profile referenced bysite-config.yaml. Falls back toenglish.mdif unset.<output.path>/— existing articles. Don't duplicate angles. Cross-link where relevant.CLAUDE.mdat project root, if present — supplemental project conventions, do-not-do list, hard rules.
If site-config.yaml is missing, STOP and tell the operator to copy
examples/site-config.example.yaml from this repo into their project
root. Don't draft against defaults — you'll write the wrong voice.
Inputs
The operator provides ONE of:
- A target keyword — e.g.
"engineering onboarding checklist". The skill picks the angle from the SERP. - A keyword + angle — e.g.
"engineering onboarding checklist, what the first-week meeting cadence should be". - No input — read the project's keyword pool if one exists (common
paths:
.claude/seo/keywords.md,content/keywords.yaml,seo/queue.md). Pick the highest-priority unwritten keyword. Confirm with the operator before drafting.
Pipeline
Run these steps in order. Each step uses the named upstream skill; inject the site-config constraints as additional system context.
Step 1 — content-brief
Invoke content-brief with:
- target keyword = the input keyword
- language = from
site-config.yamllanguagefield (defaults toen) - market = from
site-config.yamlaudience.marketfield - search intent = let the skill auto-detect from the SERP
- product context =
site-config.yamlsite.nameandsite.positioning
Capture the brief output: SERP gap analysis, recommended H2/H3 outline, target word count, must-cover entities.
Step 2 — write-content (CORE)
Invoke write-content with the brief from Step 1, plus these
config-driven overrides:
Language and register:
- Language: from
site-config.yamllanguage(defaulten). - Register: from
audience.register. Honor the value verbatim — do not paraphrase to a "nicer" register. - Pronouns: from
voice.pronouns. - Tone: must hit every adjective in
voice.tone. Must avoid every item invoice.avoid.
Terminology rules (HARD):
- Apply every
terminology.preferred"use X, not Y" pair. - Reject every entry in
terminology.bannedfrom the draft. If the topic genuinely requires a banned term, STOP and ask the operator before continuing.
Pricing rules (if pricing block present in config):
- Pricing source of truth =
site-config.yamlpricingblock. - Never invent prices. Never reference legacy prices.
Phasing dates (if phasing_dates block present in config):
- Phasing dates source of truth =
site-config.yamlphasing_dates. - Always quote the full sequence when relevant. Never paraphrase to a single date when the reality is staged.
- This block exists for regulatory rollouts (KSeF, GDPR, accessibility acts, etc.) where misquoting a date is a liability.
Brand positioning:
- Use
site.positioningonce, naturally, in the article. Not a sales pitch. One paragraph, mid-article. - Closing CTA: from
cta.textlinking tocta.url.
Article structure (defaults — config can override):
- Word count: 1,500–2,500 (write-content's default range).
- Lead paragraph (~80 words) framing the reader's question/problem.
- 4–7 H2 sections; H3 for sub-points where helpful.
- Concrete examples relevant to
audience.primary. - One natural product mention per article. Not a sales pitch.
Anti-AI-slop ruleset:
The write-content skill's built-in anti-slop ruleset (banned
vocabulary, structural pattern detection, the Horoscope Test) is
mandatory. Do NOT skip it. The voice.avoid list in site-config is
ADDITIONAL, not a replacement.
Language profile constraints: Apply every rule from the active language profile (loaded in pre-check step 2). These are language-specific style rules: punctuation density, diacritics, plural agreement, smart quotes, surface-conditional rules.
Step 3 — eeat-audit
Invoke eeat-audit on the draft. Check for:
- Experience signals: real, specific examples for the target audience. Not generic.
- Expertise: correct technical details. Domain terms used accurately.
- Authoritativeness: primary-source links where the draft makes claims of fact (statutes, statistics, official documentation).
- Trustworthiness: disclaimers where appropriate (legal/medical/ financial topics). Source attribution.
Apply suggested fixes inline.
Step 3a — Regulatory fact-check (CONDITIONAL HARD GATE)
If site-config.yaml has regulatory_topic: true (e.g. legal, tax,
medical, financial, compliance content), audit the draft for:
- Dates — every date claim must match
phasing_datesexactly OR cite a primary source. Paraphrasing to a single date when the reality is phased is a must-fix. - Numbers — every penalty/fine/threshold must be cited with source. No bare claims.
- Statute or schema versions — current version only. Never claim a future or deprecated version is current.
- Official URLs — verify before citing.
- Required disclaimer — if
disclaimer_templateis set in config, append it to the article. Verbatim.
If a claim cannot be verified against a citable source, EITHER cite the source or remove the claim. If neither is possible, STOP and tell the operator. Do not auto-publish content with unverifiable regulatory claims.
Step 4 — Language review
Invoke the bundled language-review agent (in
.claude/agents/copy-reviewer.md) with:
- The article draft
- The active language profile path
- The site-config terminology rules
Apply all must-fix and should-fix items. Surface nits to the operator in the final report.
Step 5 — Emit markdown file
Write the final article to <output.path>/<slug>.md with frontmatter
matching output.frontmatter_format. Slug = kebab-case of the keyword
unless output.slug_style says otherwise.
Frontmatter formats:
astro / next (default):
---
slug: <kebab-case>
title: "<full title>"
description: "<1-2 sentence summary, ~50-80 words, used as meta description>"
date: <YYYY-MM-DD — today>
dateModified: <YYYY-MM-DD — same as date on first publish>
readMin: <estimated read time, 1 min per ~200 words>
category: "<from config category list>"
author: "<from config author>"
authorBio: "<from config authorBio>"
---jekyll:
---
layout: post
title: "<full title>"
date: <YYYY-MM-DD>
categories: [<from config>]
author: "<from config>"
description: "<summary>"
---plain:
---
title: "<full title>"
date: <YYYY-MM-DD>
---Article body: plain markdown only. No JSX. Cross-references to other
posts use path-based links ([text](/blog/other-slug)), not hash
fragments — Google doesn't index hash-only URLs as separate pages.
Step 6 — Report
Tell the operator:
- Slug + filename created.
- Target keyword + detected intent.
- Word count.
- Any nits the language-review agent flagged that weren't auto-fixed.
- Any claims the eeat-audit step flagged for human verification.
- Suggested cross-links to existing posts (if found).
Do NOT auto-commit. The operator reviews the .md file and commits when ready.
What to do when stuck
If the keyword has no clear search intent for the configured audience (e.g. consumer queries when targeting B2B), STOP and tell the operator. Don't write thin content to fill quota — better to write 0 articles than 1 that erodes brand.
If write-content returns content that violates hard rules
(terminology, banned vocabulary, language profile, phasing dates) and
a single regen doesn't fix it, STOP and ask. Don't mechanically rewrite
— the input might need adjusting.
If site-config.yaml is incomplete (missing voice.tone, missing
audience.primary, etc.), STOP and ask the operator to fill it in
before drafting. Drafting against defaults produces generic output that
will read like exactly what it is.
Caching
The system prompt for the drafting step uses
cache_control: ephemeral — sequential calls in one session reuse the
prefix and cost ~10% of an uncached call. Keep multi-article batches in
a single session to benefit.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

