How No AI Slop Preserves the Writer's Personal Voice: Technical Architecture Explained

No AI Slop preserves the writer's personal voice through a systematic editing pipeline defined in SKILL.md that profiles voice-defining traits before editing, applies only the minimum effective changes, and validates preservation through a post-edit checklist in eval.md.

No AI Slop is an open-source editing skill for ChatGPT and Codex agents that eliminates generic AI-generated patterns while protecting authorial identity. According to the petergyang/no-ai-slop source code, the system employs a rigorous voice-preservation architecture codified in specific skill definition files to ensure the output remains authentically human.

The Voice Preservation Pipeline

The core mechanism operates through a three-phase workflow implemented in SKILL.md: voice profiling, minimal intervention, and validation.

Profiling Voice-Defining Traits

Before any modification, the skill analyzes the full draft to detect vocabulary, cadence, bluntness, humor, uncertainty, digressions, and level of polish (lines 26-27 of SKILL.md). These characteristics are treated as immutable "voice traits" that must be preserved throughout the editing process. By establishing this baseline, the system creates a protective boundary around the author's natural expression.

Enforcing Minimum Effective Edits

The system adheres to the rule "Make the minimum effective edit" (line 27 of SKILL.md). This constraint forces the engine to excise only identified AI patterns while leaving the surrounding text untouched, preventing the over-polishing that typically flattens a writer's idiosyncratic style. The engine cannot rewrite entire paragraphs; it must surgically remove slop while maintaining the original sentence structure and flow.

Protecting Authorial Edge and Specificity

Beyond general style, the skill safeguards specific elements of writer personality and factual detail that generic AI replacements often strip away.

Preserving Useful Edge and Character

Line 41 of SKILL.md explicitly instructs the system to "keep strong opinions, blunt language, humor, profanity, self-interruptions, and honest admissions when they belong to the writer." The engine never substitutes safer phrasing for these edge-case expressions that define personal voice. This rule ensures that authentic roughness and personality remain intact even when removing surrounding AI-generated filler.

The Portability Test and Specific Facts

The "Portability test" (line 36) and "Protect the specific fact" (line 38) directives ensure concrete details remain intact. By avoiding generic filler and preserving specific facts, the system maintains the writer's unique perspective and personal experience within the text. The Portability test checks whether a phrase could belong to any writer or only this specific author; if it fails, the original phrasing is protected.

Validation Through the Evaluation Checklist

After editing, the system runs verification checks defined in eval.md (lines 9-12). This checklist explicitly asks: "Does the edit preserve the writer's distinctive vocabulary, cadence, bluntness, humor, uncertainty, digressions, and level of polish?" If any check fails, the draft undergoes re-editing until the voice is fully retained. This validation loop ensures that no AI slop removal comes at the cost of authorial authenticity.

Command Usage Examples

The skill is invoked through specific commands that trigger the voice-preservation pipeline:

/no-ai-slop (your draft)

This command processes the draft through the full pipeline, returning the edited text with a "What changed" summary while preserving the author's voice.

/no-ai-slop is this slop? (your draft)

In detect mode, the skill identifies all AI patterns without rewriting the text, allowing writers to review potential issues while keeping their original voice intact for manual revision.

Summary

• No AI Slop profiles voice-defining traits—vocabulary, cadence, humor, and edge—before making any changes, as specified in SKILL.md lines 26-27.
• The "minimum effective edit" rule (line 27) ensures only AI patterns are removed, not authorial quirks or natural phrasing.
• The system explicitly preserves strong opinions, profanity, and self-interruptions per line 41 of SKILL.md.
• The "Portability test" (line 36) and "Protect the specific fact" (line 38) prevent replacement of specific details with generic filler.
• Post-edit validation in eval.md (lines 9-12) verifies voice preservation before finalizing output.

Frequently Asked Questions

What specific voice traits does No AI Slop detect before editing?

According to lines 26-27 of SKILL.md, the system detects vocabulary choice, cadence, bluntness, humor, uncertainty, digressions, and level of polish. These traits form a voice profile that constrains the editing process to ensure only AI-generated patterns are targeted while the author's natural style remains untouched.

How does No AI Slop prevent over-editing?

The "Make the minimum effective edit" rule in line 27 of SKILL.md constrains the engine to cut only identified AI-generated patterns, leaving the writer's natural phrasing and stylistic choices untouched. This prevents the comprehensive rewriting that typically erases personal voice.

Can No AI Slop handle edgier writing styles with profanity or strong opinions?

Yes. Line 41 of SKILL.md explicitly requires the system to keep strong opinions, blunt language, humor, profanity, self-interruptions, and honest admissions when they belong to the writer. The skill never sanitizes authentic edge into generic "safe" phrasing.

How does the system verify that the writer's voice is preserved?

After editing, No AI Slop runs the evaluation checklist in eval.md (lines 9-12) that verifies preservation of distinctive vocabulary, cadence, bluntness, humor, uncertainty, digressions, and polish. Failed checks trigger automatic re-editing until the voice is fully retained.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}