/voice-learn

name: voice-learn
description: Run learning pass on manually edited text to improve voice profile
arguments:
  - name: file
    description: Path to the manually edited final version
    required: true
  - name: --profile
    description: Voice profile to update
    required: true

/voice-learn Command

Capture manual edits and learn voice patterns from them.

Usage

# After manually editing generated text
/voice-learn final-draft.md --profile myvoice

What This Does

1. Saves post-edit snapshot 2. Loads matching pre-review and post-review snapshots 3. Compares post-review vs post-edit (your manual changes) 4. Categorizes edit patterns 5. Checks accumulator for matching prior patterns 6. Proposes register/rule updates for patterns with evidence 7. Holds patterns below threshold in accumulator 8. Applies user-approved updates to profile files

When To Use

Run this AFTER you've manually edited a piece that was generated with /voice-generate. The learning system needs the comparison between what the pipeline produced and what you actually wanted.

Learning Over Time

Each time you run /voice-learn, the system accumulates evidence. Patterns need 3+ instances before becoming rules. This prevents one-off preferences from polluting your profile.

Skill Invocation

Skill(scribe:voice-learn)