Voice Review
Run parallel prose and craft reviewers on draft copy against a voice register before you ship or publish marketing content.
Overview
Voice-review is an agent skill most often used in Ship review (also Launch content, Grow content) that runs parallel prose and craft reviewers against a voice profile before publishing.
Install
npx skills add https://github.com/athola/claude-night-market --skill voice-reviewWhat is this skill?
- Parallel dual-gate review: prose reviewer (AI patterns, banned phrases, voice drift) and craft reviewer (naming, destina
- Hard failures such as banned phrases and em dashes are auto-fixed before advisories are shown.
- Five Required TodoWrite checkpoints from text loaded through advisories presented.
- Loads generated text, active voice register, and banned phrase list before dispatching agents.
- Advisory findings return as tables; only hard fails are corrected automatically.
- 2 parallel review agents (prose and craft)
- 5 Required TodoWrite checkpoints in the workflow
Adoption & trust: 1 installs on skills.sh; 304 GitHub stars; 3/3 security scanners passed (skills.sh audits); trending (+100% hot-view momentum).
What problem does it solve?
You have generated marketing or editorial copy that may sound generic or drift from your voice register right before publish.
Who is it for?
Solo builders using Athola Scribe voice pipelines who want a structured pre-publish gate on `.md` and `.txt` drafts.
Skip if: Teams publishing without a voice register, banned phrase list, or upstream voice-extract and voice-generate setup.
When should I use this skill?
Checking generated content for AI patterns and voice drift before publishing.
What do I get? / Deliverables
After review, hard voice violations are auto-fixed and you receive advisory tables to approve before shipping or distributing the text.
- Auto-corrected hard-fail text
- Advisory prose and craft review tables
Recommended Skills
Journey fit
Spans multiple journey phases - primary shelf plus alternate fits below.
Canonical shelf is Ship review because the skill gates publication quality on generated markdown or text, matching pre-release editorial checks. Review subphase covers dual-gate prose and craft evaluation, auto-fixes for hard failures, and advisory tables for human decisions.
Where it fits
Run dual reviewers on a changelog draft and auto-fix em dashes before merging to main.
Re-review lifecycle email variants for AI patterns after a voice-generate refresh.
Check product docs markdown for craft issues before publishing to your docs site.
How it compares
Structured dual-agent editorial checker, not a one-shot 'make this sound better' rewrite prompt.
Common Questions / FAQ
Who is voice-review for?
Voice-review is for indie content operators running Scribe voice profiles who need systematic prose and craft QA before release.
When should I use voice-review?
Use it in Ship review before merging or publishing copy; at Launch when finalizing SEO or distribution posts; and in Grow when refreshing lifecycle or support content that must match your voice register.
Is voice-review safe to install?
It reads project text and voice config; check the Security Audits panel on this page and review scribe dependencies before enabling auto-fixes on production drafts.
Workflow Chain
Requires first: scribe voice extract, scribe voice generate
SKILL.md
READMESKILL.md - Voice Review
# Voice Review Skill Dispatch dual review agents and present unified findings. ## Method: Parallel Dual-Gate Review Two agents run in parallel on the generated text: 1. **Prose reviewer**: AI patterns, banned phrases, voice drift 2. **Craft reviewer**: Naming, destinations, dwelling, devices, anchoring Hard failures (banned phrases, em dashes) are auto-fixed. Everything else returns as advisory tables for user decision. ## Required TodoWrite Items 1. `voice-review:text-loaded` - Generated text read 2. `voice-review:register-loaded` - Voice register loaded 3. `voice-review:agents-dispatched` - Both reviewers launched 4. `voice-review:hard-fails-fixed` - Auto-corrections applied 5. `voice-review:advisories-presented` - Tables shown to user ## Step 1: Load Context Read: - The generated text (from file or clipboard) - The active voice register - The banned phrases list ## Step 2: Dispatch Review Agents Launch both agents in parallel: ``` Agent(prose-reviewer): - text: {generated_text} - register: {register_content} - banned_phrases: {banned_list} Agent(craft-reviewer): - text: {generated_text} - register: {register_content} ``` ## Step 3: Process Results ### Hard Failures Apply all auto-fixes from prose reviewer silently: - Remove/replace banned phrases - Replace em dashes with appropriate punctuation - Rewrite negation-correction patterns Report: "Fixed N hard failures (X banned phrases, Y em dashes, Z patterns)" ### Advisory Tables Present both tables to the user: **Prose Review Advisories:** | # | Line | Pattern | Current | Proposed fix | |---|------|---------|---------|--------------| **Craft Review:** | Dimension | Rating | Notes | Proposed improvement | |-----------|--------|-------|---------------------| ## Step 4: User Decision For each advisory row, user can: - **Accept** (a): Apply the proposed fix - **Reject** (r): Keep the current text - **Rewrite** (w): Apply a custom fix Present as: ``` [1] Prose: Frictionless transition at "Furthermore, the..." Proposed: Cut transition, start mid-thought [a]ccept / [r]eject / re[w]rite? ``` ## Step 5: Apply Decisions - Apply accepted fixes to the text - Skip rejected items - For rewrites, incorporate user's version - Save final text ## Step 6: Snapshot (if learning active) If the user has learning mode enabled: - Save "post-review" snapshot (text after hard-fail fixes, before user decisions on advisories) - Save "post-fixes" snapshot (text after user decisions) - Both go to `~/.claude/voice-profiles/{name}/learning/snapshots/` ## Integration with voice-generate When dispatched from voice-generate, the flow is: 1. voice-generate produces text 2. voice-generate calls voice-review 3. voice-review dispatches agents, processes results 4. User makes decisions on advisories 5. If learning mode: snapshots saved for later comparison ## Standalone Usage Can also be run on any existing text: ``` /voice-review path/to/file.md --profile myvoice --register casual ``` ## Verification After the review completes, validate these conditions: - Both review agents returned results (no timeouts) - Hard failures auto-fixed and diff shown to user - Advisory tables presented with accept/reject/rewrite options - User decisions applied to the final text - Final text saved to disk - Snapshots saved (if learning mode active) ## Test Spec The test suite (`test_voice_review.py`) validates: - Skill file exists and references parallel dispatch - Hard failure vs advisor