trove-unslop

<!-- AUTO-GENERATED from SKILL.md.tmpl — do not edit directly --> <!-- Regenerate: bun run build:skills -->

Session Init

This skill ships Trove conventions. Prefer existing project patterns over generic best practices when they conflict.

trove-unslop

Cut the tells that mark writing as AI-generated, without dropping content. Write it clean the first time; a cleanup pass after the fact reliably misses things.

Process

1. Scan for the patterns below.
2. Rewrite each into plain, direct prose. Short declarative sentences, one thought each.
3. Add voice. Sterile, voiceless writing is just as obvious a tell as purple prose. Say the real thing.
4. Self-audit. Ask: "What in this still reads as obviously AI-generated?" Fix that.

The bans we keep catching

• No em dash, and no parentheses used as its substitute. Make it two sentences, or use a colon before a list.
• No colon as a mid-sentence connector. A colon before a list is fine; a colon joining two clauses is not.
• No "It's not just X, it's Y" / "X isn't about Y" constructions. State the thing.
• No hedging stacks ("it's worth noting that", "generally", "in many cases") when you can just say it.
• No empty openers ("In today's fast-paced world", "When it comes to").
• No section-header-plus-dash one-liners. Write the header as its own sentence.

The full catalogue of patterns with fixes is in references/patterns.md.

Scope

This applies to any prose surface you produce: the reply to the user, PR descriptions, release notes, specs, commit bodies, and docs. It does not rewrite quoted source material, code identifiers, or a user's own words.