v1.1.1

humanizer 1.1.1

Patch release. Documentation consistency fixes only. No change to the de-slop and voice-match method or behavior.

Fixed

• Worked-example count corrected from three to four in SKILL.md and references/examples.md (the stance-mode example, Example 4, had been added without updating the count).
• AI-vocabulary tell citation in references/examples.md Example 1: pattern 15 (diff-anchored writing) corrected to pattern 16 (overused AI vocabulary).

Added

• Keep a Changelog version-comparison links in CHANGELOG.md.

Full changelog: v1.1.0...v1.1.1

v1.1.0

humanizer 1.1.0

Minor release. Adds 5 more tool integrations (8 to 13 supported tools) and corrects internal documentation drift. No change to the de-slop and voice-match method itself.

Added

• Five more tool integrations, bringing supported tools from 8 to 13: Windsurf (.windsurfrules), Cline (.clinerules), Continue and Zed (.continue/rules/humanizer.md), and Aider (CONVENTIONS.md). Every adapter points at the same SKILL.md and references/, so the workflow is identical across tools.
• "Where this comes from" section in the README crediting Scriveno, the longform writing system this skill's voice-preservation logic is drawn from.

Fixed

• Restraint example (Example 3) described "two em dashes" the text never contained and cited pattern 21; reframed it around the colon, parenthetical, and tricolon actually present, with the citation corrected to pattern 22.
• Aligned the eval #3 description with the corrected example.
• Synced the compatibility field in SKILL.md to all 13 supported tools.
• Corrected the tool name from Scriven to Scriveno in SKILL.md and references/voice-matching.md.

Full changelog: v1.0.0...v1.1.0

v1.0.0

First stable release.

humanizer is a pure-prompt skill: instructions only, no scripts, no
dependencies, no network. It rewrites AI-sounding prose so it reads as
genuinely human, and rewrites in a specific writer's voice when a sample or
profile is available.

What it gets right that most de-slop tools do not:

• Faithfulness over liveliness. A three-layer guard plus a mandatory meaning
  check stop the rewrite from inventing facts, causes, or quotes to sound
  concrete.
• Restraint by design. A density pre-check scales effort to evidence, and an
  explicit do-not-flag reference protects already-human writing instead of
  flattening it.
• Variance, not synonym-swapping. It targets the real signature of machine
  text, uniform rhythm, rather than relocating it with a thesaurus.
• Voice-first, with an opt-in stance mode for edge on request.

Ships with a 32-pattern tell catalog (six families), four worked examples, a
verification eval set, and adapters for Claude Code, Cursor, Codex,
Antigravity, Gemini CLI, Pi Coder, OpenCode, and GitHub Copilot.

See CHANGELOG.md for the full list.