v1.3.1

Documentation and help-text release — no behavior changes.

• SKILL.md restructured: commands grouped by operation, with humanize (whole-document, report-targeted, and the free second pass) front and center
• Full value lists are now documented for --style (12 citation styles) and --to / --source-lang (12 languages), both in SKILL.md and in --help
• Skill trigger description rewritten to lead with humanizing and state format support precisely (humanize takes .docx/.pptx; translate additionally takes .pdf/.xlsx/.epub/.html/.txt)
• README tagline is now "Humanize what's flagged. Preserve the rest."
• --min-words/--max-words help now says the bound describes the rewritten text, not the source

v1.3.0

What's new

New command free-rehumanize <job_id> --report new.pdf — continue a finished humanize job for free. Only the passages a fresh report still flags on that job's result are rewritten, at no credit cost.

The server enforces every limit: one free continuation per job, a per-day cap, and the report must match that job's result (≥90%) and show ≥20% AI, with the result still retained (~7 days). --segments-file optionally bounds the flagged passages per passage. SKILL.md now points the "second pass" at this free path.

v1.2.0

What's new

humanize gains word control, matching the app's two paths:

• Whole document — --min-words / --max-words hold the whole rewrite to a word range.
• Per-passage — --segments-file budgets.json (a list of {text, min_words?, max_words?}) bounds individual flagged passages. Get the passage texts from the report command and pass --report alongside so it defines the scope.

Word control is experimental: constraining word count noticeably weakens AI-rate reduction, so leave it off unless a length is required. The whole-document band cannot combine with --report or --segments-file — use one path or the other.