write-post

Write and validate documentation portal articles with conclusion-first structure and verifiable numbers.

Updated Aug 6, 2021
One-click install
npx skills add https://github.com/the-homeless-god/dotfiles --skill write-post-the-homeless-god
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: write-post
Source: https://github.com/the-homeless-god/dotfiles/tree/main/skills/write-post
Command: npx skills add https://github.com/the-homeless-god/dotfiles --skill write-post-the-homeless-god

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Articles on documentation portals often get rejected as "not written for humans" or "AI slop" because they bury the answer, invent unsearchable terms, hide what was not measured, and print numbers no command can reproduce. This Skill encodes six editorial rules, learned from four real rewrites of one article, that make a post readable, searchable, and falsifiable. ## Core Features & Use Cases - Conclusion-first structure: Enforces a shape where the title states today's conclusion, the first screen carries the answer with numbers, and the author's post-mortem moves to the end. - Terminology and honesty rules: Requires accepted searchable terms glossed once at first use, a mandatory section naming what was never measured, and negative results shown beside positive ones. - Verifiable numbers: Every number must name the command or file that prints it, and a checker script (scripts/check-post.sh) validates shape, first-screen caps, term glosses, and border sections. - Use Case: Rewriting a study write-up that was returned as "a small article about nothing" — restructure it so the answer leads, all nine measurements appear, and an 11-item "what we did not measure" section closes the knowledge boundary. ## Quick Start Ask the agent to write or repair a documentation portal article using the write-post rules, then run scripts/check-post.sh on the resulting file to verify its structure.

Frequently Asked Questions about write-post

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I structure a technical article so readers find the answer immediately?

Put the conclusion in the title, place a table of answers with numbers above the first section heading, and move any post-mortem or correction history to the end. The reader who stops after the first screen should still leave with the answer and one line of what to do with it.

How to write a study write-up that does not get rejected as AI slop?

Show all measurements including negative results in one table, add a section naming what you never measured, and make every number reproducible by naming the command that prints it. Avoid inventing terms; use accepted spellings like baseline or few-shot with a one-line gloss at first use.

What does the check-post.sh script validate in an article?

It checks that the file has one H1, a first screen within its character cap carrying numbers, a knowledge-border section, a named negative result, something copyable, and every Latin term explained at first use. It validates shape only, never truth, and --selftest shows it failing on eight prepared bad cases.

When should I not use this article-writing approach?

Do not use it for a tool's own documentation section, for reference material generated from source code, or for status reports. Those formats have different structural rules and dedicated skills cover them.

Why must every number in an article name a command?

A number no command prints cannot be refuted, so it ages silently after the underlying system changes. Naming the command, such as wc -l or grep -c, lets any reader rerun the measurement and catch drift between the article and reality.