sanitize-artifacts

Removes prompt leakage and production residue from generated artifacts.

Updated Jul 16, 2026
One-click install
npx skills add https://github.com/Kotomiya07/skills --skill sanitize-artifacts-kotomiya07
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: sanitize-artifacts
Source: https://github.com/Kotomiya07/skills/tree/main/plugins/sanitize-artifacts/skills/sanitize-artifacts
Command: npx skills add https://github.com/Kotomiya07/skills --skill sanitize-artifacts-kotomiya07

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Artifacts produced through iterative prompting often leak conversation history, prompt instructions, constraints, and corrective feedback into the final deliverable, making it read like a patchwork of the production process rather than a coherent document for its real audience. ## Core Features & Use Cases - Leakage Detection: Identifies sentences that expose prompts, constraints, avoided approaches, tool choices, or meta-commentary that the audience does not need. - Constraint-to-Design Conversion: Rewrites explicit production rules (e.g., "do not use Git") into natural artifact design (e.g., Drive-based workflow instructions) instead of visible disclaimers. - Coherence Review: Applies a 10-point inspection checklist covering tone consistency, single voice, audience-appropriate headings, and removal of patchwork from multiple feedback rounds. - Use Case: After vibe-coding a user guide through several rounds of corrective feedback, run this skill to strip out phrases like "As requested" or "This avoids using CLI tools" so the guide reads as an intentional, standalone deliverable. ## Quick Start Review the guide we just wrote and sanitize it so it reads as a standalone deliverable with no traces of our conversation or instructions.

Frequently Asked Questions about sanitize-artifacts

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I remove prompt leakage from AI-generated documents?

Review the artifact for phrases like "As requested", "Based on your instruction", or explicit mentions of constraints and avoided approaches. Rewrite these as natural design decisions so the document reads as intentionally authored for its audience.

How to clean up artifacts created through iterative prompting?

Classify each statement as audience-needed content, production guidance, or conversation residue. Keep only the first category directly, reflect the second through structure and tone, and delete the third to eliminate patchwork from feedback rounds.

Should constraints from the conversation appear in the final deliverable?

Constraints should normally shape the artifact's design invisibly rather than appear as disclaimers. For example, instead of stating "This guide does not use Git", write Drive-based workflow instructions that apply the constraint without exposing it.

When should examples from the conversation be kept in the artifact?

Keep a conversational example only when the artifact itself specifically needs it. Examples given to communicate intent are diagnostic material; use them to infer tone, audience, and abstraction level rather than copying them into the deliverable.

What are the limitations of automated artifact sanitization?

Sanitization requires judgment about what the audience genuinely needs, so mechanical removal can strip necessary context. The skill preserves technical correctness and user-facing requirements, but borderline disclaimers still need human review.