| name | anti-slop-slop-canon |
| description | Write, compose, rewrite, edit, or audit everyday text such as notes, updates, posts, emails, docs, scripts, narration, and voice-agent replies. Applies your voice without AI-writing habits. Also covers voice onboarding or setup, profile inspection or refresh, and realtime style-prompt requests. |
Anti-Slop Slop Canon
Use default content version 0.1.0.
Resolve scope and one bundle
- In a project copy, inspect only
<project-root>/.anti-slop-slop-canon/; never inspect or fall back to global state. Otherwise inspect only ~/.config/anti-slop-slop-canon/.
- Validate
voice-profile.md with references/profile-schema.md. Require schema 1.0.0, active scope, every required section, and actionable rules. Load it alone; otherwise load references/defaults.md alone. Report invalidity after the task and offer a previewed repair.
- When profile and
settings.md are absent, gate the first natural-language task: read references/onboarding.md and offer personalize, defaults, or defer before writing. Also read it for explicit onboarding, setup, or voice learning; otherwise do not.
- With a profile, compare its loaded
defaults_version to 0.1.0 without reading defaults. Finish the task with that profile before any notice or lifecycle action.
Protect and route
- Select spoken only when text is clearly meant to be heard. Otherwise use written without asking.
- Preserve meaning, facts, uncertainty, intent, exact quotations, code, structured data, and legally fixed wording. Reject detector-evasion framing and offer clarity or voice editing.
- Read references/operations.md. Route explicit operations and lifecycle choices; infer compose for creation and rewrite for revision.
- Apply one bundle and mode. Silently check meaning, exemptions, mode, and rules, then fix only mutable prose.
- Return clean compose or rewrite output. Add reasoning or a diff only when requested. Audit without editing.