Sound like yourself.
anti-slop-slop-canon is an open-source Agent Skill that removes recognizable AI-writing habits from your prose, then learns your actual voice through a short, adaptive onboarding session. One profile governs both written and spoken output — docs, posts, emails, scripts, narration, and voice-agent replies. It runs on the model and tools your host agent already has, with no runtime package, API key, hosted inference, or telemetry.
Install
npx skills add codeSTACKr/anti-slop-slop-canonThe installer asks whether you want global or project scope. Works with agents that read Agent Skills, including Claude Code, Cursor, and Codex.
What's in 0.1.0
- Default writing canon — removes hype, marketing jargon, em/en dashes, semicolons, exclamation points, mid-sentence colons, rule-of-three, "here's the thing," and manufactured contrast, without inventing claims the source didn't make.
- Adaptive onboarding — one-question-at-a-time voice learning with preview-and-approve before anything is saved, and no retention of samples or analysis by default.
- Profile lifecycle — read-only inspection, respected direct edits, version-mismatch detection, and approval-gated refresh.
- Realtime voice prompt — a compressed style module you can paste into a voice agent's system prompt.
- Public showcase — useantislop.com, where every example is real output from the bundled rules.
See CHANGELOG.md for full detail.