Clarity 0.2.1

This patch keeps the Clarity site dependencies current and secure.

• Upgrades the site to Astro 7.3.1 and Node 22
• Clears the current npm security audit
• Keeps the skill guidance and editor behavior unchanged

clarity 0.2.0

What's new since 0.1.0

• A site to go with the skill. clarity.addy.ie now has a writing editor, a page explaining how Clarity works and how it's evaluated, and a walkthrough comparing the three sample drafts side by side.
• Built for agents, not just browsers. llms.txt, .well-known/agent-instructions.md, and content-negotiation edge functions make the site's content directly consumable by agents, not just readable in a browser.
• A smaller, more focused runtime for the skill itself.
• Easier to find and navigate — improved social previews, a refreshed editor interface, and clearer navigation for both people and agents.

clarity 0.1.0

An Agent Skill that drafts, rewrites, and reviews prose so it carries a point of view instead of reading like generic model output.

npx skills add addyosmani/clarity

Then pick a mode:

/clarity interview <topic>    it interviews you, then co-writes from what you said
/clarity rewrite draft.md     it edits a draft you already have
/clarity review draft.md      it critiques and leaves your file alone

What is in it

• Eighteen rules for writing that earns its reader, with the reasoning behind each one.
• An interview, not a briefing. On an empty page the skill asks you questions before it writes anything. Context re-worded by a model reads as generated; your own sentences do not, so it collects the latter and arranges them.
• Earned-or-unearned judgement instead of blanket bans. A rhythmic series of concrete nouns, an earned closing beat, and a cumulative sentence that keeps its subject all survive the edit pass. Sweeping rules against adverbs, passives, and dashes do not survive testing.
• A register gate. Only an argument owes the reader a disputable thesis. Demanding one from a piece about wonder produces an argument against wonder, which is how the first draft of the sample essay went wrong.
• Two axes of concrete. Verifiable specificity makes a claim checkable; sensory concreteness makes a reader see something. Most drafts need the second more, and a rewrite that only knows the first turns into a bibliography.
• A diagnostic linter that deliberately reports no composite score, because a stylometric composite inverts on real authors.

Samples

samples/ holds two essays before and after a pass, and a third built from an interview with the author, shipped alongside its transcript. The interviewed version is the one worth reading: roughly 87% of its three-word sequences come straight from what he said out loud.

Known limits

The rules in the README are the author's own and are quoted verbatim by the skill. Two of them, on stating an arguable point and on specificity, predate the register gate and the sensory axis and are narrower than the skill now behaves.