Humanize Redo

Re-runs the strict revision phase of a Korean humanizing pipeline on the latest run, optionally scoped to a category, paragraph, or intensity.

What it does

This skill locates the most recent _workspace/{run_id}/ directory in the current working directory and re-invokes the strict revision phase (Phase B) of the humanize-korean skill on it. It parses the user's instruction to limit re-revision to a specific category of findings, a specific paragraph, or a chosen intensity level, or to roll back a particular edit. Outputs are versioned as 03_rewrite_v2.md (or v3), with the previous final.md backed up as final_prev.md, followed by parallel verification and a final comparison table with new grades. It caps at round 3 and recommends human review beyond that.

When to use it

Use it when a previous humanize-korean run left residual findings that need a second pass, or when only part of a document needs re-polishing. It is also used to adjust revision intensity or roll back a specific change. It requires an existing prior run; otherwise it tells the user to start with /humanize.

Requirements

Runs inside an agent with Glob access to the working directory and depends on the humanize-korean skill and its components (korean-style-rewriter, content-fidelity-auditor, and the ai-tell-taxonomy reference). It ships no scripts of its own and needs no credentials or network access.