Your Rules, Enforced Every Time
Not AI opinion. Not suggestions that change between runs. Deterministic prose rules that you define, scoped to your project, your documents, your scenes, your characters. The same input always produces the same output.
5 Rule Categories
| Category | What It Checks |
|---|---|
| Structural | Sentence length variance, paragraph breaks, dialogue-to-prose ratio, scene pacing metrics |
| Stylistic | Adverb density, passive voice, repeated words, filter words, purple prose detection |
| Semantic | Cliche detection, mixed metaphors, tone shifts, anachronistic language |
| Continuity | Character name consistency, timeline violations, setting detail contradictions |
| AI-isms | Detects patterns common in AI-generated text — hedging language, excessive qualifiers, formulaic transitions |
Scope Hierarchy
Rules cascade through four levels: Project, Document, Scene, and Character. Set a project-wide rule against passive voice, then override it for a specific character's internal monologue. The scope hierarchy means you never have to choose between consistency and flexibility.
Rule Editor
Configure every rule with precision. Set thresholds (max adverb density: 3%), choose severity levels, define exceptions. The editor shows live previews of how each rule affects your current document.
Scope Hierarchy
Rules cascade from project-wide defaults down to individual character overrides. Your villain can use passive voice. Your narrator follows strict style rules. Same project, different rules, zero conflict.
Severity Levels
Four levels — error, warning, info, hint. Configure which issues block your flow and which are gentle nudges. Your project, your thresholds.
YAML Import/Export
Define rulesets as YAML files. Share them between projects, distribute them to writing groups, version-control them in git. Your style guide, codified.
ProseGuard vs. generic grammar tools
| Feature | Grammarly / ProWritingAid | Ishvana |
|---|---|---|
| Rule enforcement | AI suggestions that change each run | Deterministic rules — identical results every time |
| Character voice | Flags dialect as errors | Scoped per character — each voice has its own rules |
| Scope control | Global settings only | Project > Document > Scene > Character hierarchy |
| Custom rules | Limited preset categories | Full YAML rule definitions, import/export between projects |
| AI text detection | Not available | Built-in AI-isms category catches machine patterns in your prose |
AI-Powered Analysis
AI-isms Detection
The AI-isms category uses pattern analysis to detect machine-generated text characteristics — hedging language ('it's worth noting that'), excessive qualifiers, formulaic paragraph transitions, and the telltale smoothness of LLM prose. Useful whether you're checking AI-assisted drafts or ensuring your own writing doesn't accidentally read like ChatGPT.
Pre-Generation Hints
ProseGuard feeds your active rules to Hawken before it generates text. If your rules say 'no passive voice in action scenes' and you're writing an action scene, Hawken knows before it writes a single word. Prevention, not just detection.