SideButton
W

Knowledge Pack Files

Writing Standards Knowledge Pack Files

Browse the source files that power the Writing Standards MCP server knowledge pack.

Available free v1.0.0 LLM
$ sidebutton install writing
Download ZIP
writing-quality/ATTRIBUTION.md
2.1 KB

Attribution

This module merges and adapts content from two MIT-licensed open-source projects:

humanizer

  • Repository: blader/humanizer
  • Author: Siqi Chen
  • License: MIT (Copyright 2025 Siqi Chen)
  • Contribution: 29 AI writing pattern definitions, two-pass audit process, voice calibration methodology
  • Source: Based on Wikipedia's "Signs of AI writing" page maintained by WikiProject AI Cleanup

stop-slop

  • Repository: hardikpandya/stop-slop
  • Author: Hardik Pandya
  • License: MIT (Copyright 2025 Hardik Pandya)
  • Contribution: 5-dimension scoring system (Directness, Rhythm, Trust, Authenticity, Density), 35/50 threshold, banned phrases, banned structural patterns, 8 core editing rules, 12-item quick checks

Upstream Sync

This module is structured to allow tracking of upstream changes. To sync:

  1. Compare writing-quality/_skill.md against the latest SKILL.md from each source repo
  2. Compare references/banned-patterns.md against humanizer's pattern list
  3. Compare references/banned-phrases.md and references/banned-structures.md against stop-slop's reference files
  4. Merge new patterns or rules, preserving this module's SideButton formatting

Both source repos are actively maintained. Check for new versions periodically.

Relationship to plugin-writing-quality

The sidebutton/plugin-writing-quality repo provides a standalone MCP plugin with deterministic JS pattern checking + LLM scoring. It uses the upstream repos as git submodules directly.

This skill pack module adapts the same upstream content into LLM-consumable reference files with additional calibration:

  • Landing-page content-type exceptions (Pattern 10, 29, S3, Metronomic)
  • Content-type-aware scoring adjustments (Rhythm, Trust, Authenticity for landing pages)
  • Calibrated thresholds (28/50 for landing pages, 35/50 for prose)

When updating either, sync the calibration changes to the other:

  • Skill pack references → plugin's scoring-prompt.js content-type logic
  • Plugin's upstream submodule updates → skill pack reference files