Advanced Human Writing & AI Humanizer

Humanize AI-shaped text, write or continue genre-aware prose, and audit long-form continuity without flattening a specific voice. Use the smallest set of modules that covers the task. Keep project facts, prior chapters, rewrite originals, and continuity ledgers separate from optional style references.

Capability Layers

• The humanwriting/ Python package is executable. Its CLI deterministically compiles prompts, locates recurring text patterns, calculates diagnostics, previews conservative fixes, checks protected content, and writes staged audits.
• The Markdown files under skills/ are model-executed writing and editorial modules. They are selected by the compiler and are intentionally not pretending to be deterministic NLP algorithms.
• A normal installation includes both layers. Verify the executable layer with human-writing-skills list --kind module and run a real draft through lint, fix, verify, or pipeline rather than judging the package from SKILL.md alone.
• human-writing-mcp is an optional project-local coordination layer. It gives separate agents bounded long-form assignments, stores receipts, and refuses reconciliation until coverage verification passes. It does not call a model.

Quick Humanize Route

• For a supplied draft, use humanize --mode quick for surface patterns, rewrite fidelity, and voice/ambiguity preservation.
• Use --mode deep only when structural repetition, cliches, paragraph stagnation, or broad editorial reconstruction needs a separate pass.
• Load humanize-examples only after an explicit request or --with-examples.
• Do not activate rewrite-preservation modules for unrelated new drafting.

Workflow

1. Select one base style from skills/: fiction, webnovel, argumentative, news-report, formal-document, self-media, or academic-paper. Before producing prose, read skills/generation-quality.md and apply its compact protocol during drafting. The CLI build and MCP compile_generation_prompt include it automatically. Prevent repeated rhetorical moves, redundant explanation, decorative details, invented specificity, and unearned closure before they accumulate. Let genre, source evidence, and author voice govern choices; do not force roughness, body reactions, fragment runs, or open endings as proof of naturalness.
2. Read only the relevant modules. Add continuity, spatial, relationship, number, dialogue, dialogue-performance, register, capability, world, process, salience, recurrence, source, rhythm, preservation, or AI-trace modules when the text actually needs them. narrative-naturalness-audit is reserved for deep or explicit AI-trace review of narrative prose; it is not loaded for ordinary quick humanization or serious documents. For fiction with spoken or wordless interpersonal exchanges, use dialogue-voice-audit and dialogue-performance-audit (CLI: --profile voice). Check both participants and separate speech from simultaneous action; establish reception, refusal, or an evidenced pending state before assuming completion. During drafting, reconsider this route when an interaction enters the scene even if the original request only said to continue a chapter.
3. Treat user facts and --context as authoritative. Never borrow facts from a style sample. When --original is supplied for a rewrite, activate rewrite-fidelity and preserve meaning without preserving awkward wording. In fiction and webnovels, do not let time/place mini-headings replace scene bridges.
4. Activate reference-style-alignment only when the user supplies reference material, gives an explicit style direction, or directly asks to match a style.
5. Treat --source as factual evidence only. Activate source-grounding for serious documents or explicitly selected factual self-media / argumentative work with source files. Personal experience and product reviews need evidence too.
6. For important revisions, run deterministic lint, then independent audit profiles, then verify protected literals against the source. For consequential rewrites use verify-fidelity too: changed claims require review even if every number and citation is unchanged. Run stats only when distributional diagnostics help; use fix as a preview before writing.
7. Keep voice, serial, world, process, momentum, salience, recurrence, ending, texture, fidelity, preservation, examples, and sources separate from the default audit. Activate them explicitly or through pipeline --auto. serial requires context, recurrence requires at least three chapters, fidelity requires --original, preservation requires both explicit selection and --original, examples require an explicit request, and sources requires a factual document type and --source. For diagnosis without replacement prose use critique. For important revisions, compare-blind writes an A/B prompt and a separate version key: send only the prompt to a reviewer without revision history. Accept ties and original wins. Stop when further edits have no evidenced benefit; never use AI-trace scores as quality proof.
8. For a book-length manuscript or large report, use chunk-audit instead of placing the whole draft in one prompt. Supply --outline or --context for authoritative character or report rules. Use --agent-mode deep only when every block needs a visible coverage receipt, then run verify-chunk-audit before reconciliation. Use --translationese only for an explicitly translated or localized work. Read docs/long-form-consistency.md only for this workflow. Add --book-level only for an explicitly requested whole-book fiction/webnovel review with at least three identifiable chapters; run it after chunk reports.
9. When multiple agents or sessions share a long-form project, use human-writing-mcp rather than copying the entire manuscript into each chat. Agents must claim a task, submit a complete receipt, and pass coverage verification before one final reconciliation task is released. Read docs/agent-orchestration.md only when an MCP-capable host is available.
10. For requested long-form generation (typically 10,000+ output characters/words) or explicit multi-agent writing, read skills/long-form-writing.md on demand. Use plan-writing / plan_long_form_writing to check the outline, allocate bounded writer/editor tasks, and gate assembly on cross-block continuity. Dependent scenes run sequentially; explicit independent state contracts permit bounded parallel work. The host supplies native subagents; disclose role-pass fallback when no runner exists. An existing long input alone does not activate long-form generation. For the detailed commands read docs/long-form-writing.md.

Commands

human-writing-skills build --style fiction --context ledger.md --task "Continue the scene."
human-writing-skills humanize --draft chapter.md --style fiction --mode quick
human-writing-skills humanize --draft article.md --style self-media --mode deep --with-examples
human-writing-skills build --style fiction --reference sample.md --task "Match the sample's restrained rhythm."
human-writing-skills build --style self-media --original original.md --task "Rewrite without adding facts."
human-writing-skills audit --draft chapter.md --context ledger.md --profile physical
human-writing-skills audit --draft chapter.md --profile voice
human-writing-skills audit --draft chapter.md --context ledger.md --profile serial
human-writing-skills audit --draft chapters.md --profile momentum
human-writing-skills audit --draft chapters.md --profile recurrence
human-writing-skills audit --draft chapter.md --profile process
human-writing-skills audit --draft chapter.md --document-type fiction --profile ending
human-writing-skills audit --draft paper.md --document-type academic-paper --source study.md --profile sources
human-writing-skills audit --draft revised.md --original original.md --profile fidelity
human-writing-skills audit --draft revised.md --original original.md --profile preservation
human-writing-skills pipeline --draft chapter.md --context ledger.md --auto --with-stats --output-dir audit
human-writing-skills chunk-audit --draft full-novel.md --style fiction --outline outline.md --output-dir novel-audit
human-writing-skills chunk-audit --draft full-novel.md --style fiction --outline outline.md --agent-mode deep --output-dir novel-agent-audit
human-writing-skills verify-chunk-audit --package-dir novel-agent-audit
human-writing-skills plan-writing --outline outline.md --context ledger.md --style fiction --target-length 60000 --chunk-size 5000 --output-dir book-writing
human-writing-skills writing-tasks --package-dir book-writing
human-writing-skills chunk-audit --draft report-es.md --style news-report --source sources.md --translationese --agent-mode deep --output-dir report-audit
human-writing-mcp --root C:\writing-project
human-writing-skills lint --draft chapter.md --style fiction
human-writing-skills stats --draft chapter.md --style fiction
human-writing-skills fix --draft chapter.md --preview
human-writing-skills verify --source original.md --candidate revised.md
human-writing-skills verify-fidelity --source original.md --candidate revised.md
human-writing-skills evaluate --cases tests/fixtures/quality/multilingual-benchmark.json

Read README.md or README.zh-CN.md for user-facing guidance. Read files under docs/ only for the workflow being used. Do not claim detector evasion or infer authorship from stylistic patterns; frame results as editing evidence.