- Home
- Skills
- Technical Documentation
- Master Writing
Works with the AI Agents you already use
Master Writing
Prose canon for text a human reads. ASD-STE100 strict or flavored modes, plain words, no dash or space before colon, banned machine-style patterns, no invented facts, format limits, script │ check.
$1
Master Writing
Example session with this skill installed
use master writing skill to write a new blog article about "Les Grands Jeux D'Antan".
- Read your context and instructions
- Compiled the master writing
Les Grands Jeux d'Antan: building a game book for care homes in the Gers
Published 10 November 2026
Les Grands Jeux d'Antan is a printed game book for residents of French care homes, known as EHPAD. It collects memory, language, logic and social games in large type. A game works alone, in pairs or in a workshop with the activity leader. This article explains the format, price and print choices that make the book usable in a care home in the Gers. It also gives the calendar that shapes the 2026 campaign.
The ground: 36 care homes in 28 communes
The Gers has 36 care homes in the official count, spread across 28 communes. The list recalculated on 4 November 2026 gives 46 lines when attached units are included, for about 2,956 places. Thirty-nine homes are approved for social aid, and nineteen Alzheimer units are listed. Eight sit in Auch. The median monthly fee is €2,259, based on CNSA declarations from January 2026.
What the book must do in a bedroom
The book collects memory, language, logic and social games inspired by older pastimes. A game works alone, in pairs or in a workshop with the activity leader. Large type and pictograms lower the reading barrier. Paper can be handled, passed around, and needs neither a connection nor a battery. The book also works as a gift for residents, families and volunteers. Each page carries one simple rule, so no separate instruction sheet gets lost.
The A5 format fits in one hand
We chose a closed A5 format, 148 by 210 millimetres, with 64 inner pages. The 100 g/m² matte coated paper reduces glare from a ceiling light. The 300 g cover gets a soft touch laminate, which feels pleasant in the hand. The corners are rounded and the spine is perfect bound. The book fits in one hand and rests on a bedside table.
A price measured against one day of care
The target retail price runs from €9.90 to €14.90. For a director, the useful comparison is one day of accommodation. The Gers median reaches €57.48 per day for a single room, according to the OuiRetraite barometer of November 2026. A €14.90 book equals 26% of that day. A twenty-copy allocation costs €298. These figures are commercial hypotheses, to confirm with quotes and field feedback.
The local printer, and the trap of a small run
The Gers has 22 active printers and finishers. A local printer offers proximity, a workshop visit and fast correction of the proof. An online printer often offers a lower unit price, with less flexibility on small formats. The first known quote reaches €270 for ten copies, or €27 per unit, a prototype rate. The company behind it has since been struck off the register. Volume weighs more than anything else on unit cost: moving from ten to five hundred copies divides the price by about six.
The Christmas clock
The decision chain in a care home takes two to four weeks, from the first call to the purchase order. An order placed in November can still arrive before the holidays. An order placed in December becomes a January product. A campaign that starts on 9 November leaves two weeks for quotes and proof approval. Production then takes two weeks, and delivery one week, with each day of delay pushing delivery back. Sales work therefore starts during production.
No resident data
No resident data enters our records. The book addresses the care home, and the contacts we collect are professional and public. Email outreach identifies the sender and offers a way to refuse, with a record of objections. A workshop photo requires written consent, for residents and for staff. We promise neither a date nor a print run before the printer confirms it in writing.
What we do now
The 2026 campaign targets three pilot homes, chosen from the six priority sites. We skip a mass mailing to all 46 sites in the first wave. A demonstration copy and a thirty-minute workshop give a more useful return than a brochure. The first workshop decides the sequence: correct the book, choose the printer, then launch the run. No outreach email has gone out at the date of this article. Care homes in the Gers that want to test the book can write to contact@antan-editions.fr.
Connects securely to your tools. The creator never sees your data.
About this skill
Master Writing controls the prose the agent writes for a human reader: skills, docs, comments, commit bodies, error strings, emails, social posts, UI copy. Source code is out of scope.
Base standard is ASD-STE100, in two modes.
- Strict mode covers procedures, errors, tool docs, commits, and comments. Active voice, 20 words per sentence maximum, simple tenses only, no gerunds, noun clusters of 3 words maximum, one instruction per sentence, no humor.
- Flavored mode covers articles, chat, email, social text, and UI copy. Default active voice, 25 words per sentence maximum, noun clusters of 4 words maximum, limited humor.
Process is 6 steps: brief, one purpose sentence, outline by value to the reader, draft, revise in this order (cut, banned patterns, STE, words, reader, rhythm, typography), then run the check.
Hard rules, all formats and languages.
- No em dash, en dash, or hyphen used as a dash. Use a colon, a pair of commas, parentheses, or a new sentence.
- No space before : or before ;, !, ?.
- One colon per sentence at most. Between a verb and its object, no colon.
- Straight quotes, sentence case, diacritics kept, one exclamation mark maximum, no emoji by default.
Banned patterns: contrastive formulas such as X, not Y, absolutes and superlatives with no scope and no source, reveal questions, fragment stacks, adjective triads, slogan endings, hype words, engagement bait, empty politeness.
Truth: no invented fact, number, name, quote, or personal experience. An unknown detail becomes [TO CONFIRM: ...], and any piece with a placeholder counts as a draft.
Words: one term for one thing in the full piece, plain replacements (use for utilize, build for implement, many for a large number of).
Voice: informative by default. Personal or provocative when the brief selects it. Polite form by default in languages that have one, and the reply uses the form of the person addressed.
Formats: 12 profiles with shape rules, from chat (1 to 3 sentences, answer first) and email (150 words maximum, one next action) to blog (40 to 80 word intro, one idea per section) and release notes (one imperative bullet per change).
Social: hook in the first visible line, one idea per post, at least one concrete detail, end on a specific question or next step. Replies name the detail answered, add one new thing, and match the language and register. A thread stops after 3 exchanges. Platform limits come from a table: LinkedIn 3,000 characters with about 210 visible, X 280 weighted, Bluesky 300, Mastodon 500, Reddit title 300, Discord 2,000.
Check: scripts/check.py --mode --lang --platform prints error, flag, and info findings with an exit status of 0 only when clean. Fix each error, judge each flag, then read 11 items by hand, because the script cannot judge meaning or truth.
Any language: the same canon applies. references/FRENCH.md adds local detail. A translation keeps every meaning unit, number, name, link, section count, humor count, and placeholder.
Priority: system policy, always-english for the agent's own chat, an active master-override, this skill, then platform skills, which add rules but cannot remove sections 4 and 5.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Technical Documentation.
Does not apply to source code logic. Banned pattern script checks are optimized for English and French. Character limits reflect September 2026 platform standards.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 6 days ago
- Software
Python 3.10+
Permissions
File Scopes
check.py python script needed for some more automated checks the skills use