Humanizer-next

Humanizer-next is the source repository for an agent skill that removes common signs of AI-generated writing while preserving meaning, tone, and technical literals.

This repo is not a standalone runtime library. It exists to maintain canonical skill content, compile generated artifacts, validate adapters, and distribute synced outputs to multiple agent environments.

Repo role

• Canonical skill sources live under src/.
• Experimental prototypes and extraction candidates live under experiments/.
• Generated root artifacts are SKILL.md and SKILL_PROFESSIONAL.md.
• Adapter outputs live under adapters/.
• Repository guidance for agent environments lives in AGENTS.md.
• Installation and platform support guidance lives in docs/install-matrix.md.

Maintainer setup

git clone https://github.com/edithatogo/humanizer-next.git
cd humanizer-next
npm install

This setup is for maintainers working on the skill source. End-user install paths for Gemini, Antigravity, Copilot, VS Code, and other adapters are documented in docs/install-matrix.md.

Maintainer workflow

1. Update source fragments in src/.
2. Rebuild and sync generated outputs with npm run sync.
3. Validate adapters and docs with npm run validate.
4. Run the full maintainer gate with npm run lint:all, npm test, pytest, and npm run check:sync.

npm run check:sync is important for this repo shape. It verifies that generated adapter outputs are already in sync with source content and prevents drift from being merged.

Supported outputs

• Standard skill: SKILL.md
• Professional skill: SKILL_PROFESSIONAL.md
• Agents manifest: AGENTS.md
• Adapter bundles under adapters/

Current adapters include Gemini CLI, Google Antigravity, Qwen CLI, GitHub Copilot, VS Code, and related wrapper formats used by downstream tools.

What this repo is not

• Not published as an npm package.
• Not intended to be consumed as an application dependency.
• Not a general-purpose writing toolkit monorepo.

Release model

Releases package skill artifacts and adapter bundles as GitHub release assets. The release workflow does not publish to npm.

Quality gates

The repo is validated as a skill-source repository:

• Markdown, Vale, ESLint, TypeScript, and Prettier checks
• Node regression tests
• Python adapter tests
• Sync-drift verification
• Cross-platform skill distribution validation

The maintainer gates are intentionally centered on the maintained skill surface: src/, generated artifacts, adapters, docs, and validation scripts. Content under experiments/ is kept in-tree for evaluation and extraction decisions, but it is not treated as part of the primary supported skill contract.

Self-improvement track

The active conductor self-improvement track lives under conductor/tracks/repo-self-improvement_20260303/. It refreshes upstream repo data, reviews open PRs and issues, and records Adopt, Reject, or Defer decisions for candidate improvements.