post-merge-voice-gate_skill

This skill analyzes post-merge draft content to detect generator voice from outline transitions and routes fixes to the earliest source.
  • Python

109

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill willoscar/research-units-pipeline-skills --skill post-merge-voice-gate

  • SKILL.md4.6 KB

Overview

This skill inspects a merged draft for planner-style transitions, slide/navigation narration, and axis-label slash lists that were injected during merge. It produces a short PASS/FAIL report and a concrete routing plan so fixes land at the earliest responsible source. The gate is analysis-only and never edits the draft.

How this skill works

The gate scans output/DRAFT.md and outline/transitions.md for high-signal stems (planner-talk, navigation cues, and slash-list axis markers). If any trigger patterns are present it classifies the failure source: either the transitions file or only the merged draft. It then writes output/POST_MERGE_VOICE_REPORT.md with a PASS/FAIL verdict and an actionable routing plan; on FAIL it also appends a short record to output/QUALITY_GATE.md so the workspace remains debuggable.

When to use it

  • Immediately after section-merger has produced output/DRAFT.md
  • When you want to prevent merge-injected planner voice from contaminating paper voice
  • Before running any polishing pass that might accept generator-style prose
  • When outline/transitions.md was used during the merge and you need evidence-first routing

Best practices

  • Run this gate as an automated step right after section-merger to catch issues early
  • Treat findings as routing instructions—fix the earliest responsible artifact, not ad-hoc edits in the draft
  • Prefer minimal, content-bearing rewrites: turn navigation into argument bridges, slash lists into natural prose
  • If the offending string is in outline/transitions.md, rerun transition-weaver or hand-edit that file then rerun section-merger
  • Keep the gate deterministic and evidence-first: include matched excerpts and file locations in the report

Example use cases

  • A merged survey draft includes ‘To keep the chapter...’—gate FAILS and routes the fix to outline/transitions.md
  • Slash-list axis labels like ‘retrieval/index/write’ appear in the body—gate FAILS and suggests a prose rewrite in outline/transitions.md
  • Transition file is clean but the draft still contains slide narration—gate FAILS and routes to writer-selfloop or subsection-polisher depending on location
  • No planner-talk stems found—gate PASSES and records a clean report for the quality log

FAQ

High-signal stems such as planner-talk transitions (e.g., “To keep the chapter…”), slide/navigation narration (“Next, we move…”), and slash-list axis markers using ‘/’. If any appear in the merged body or transitions file, the gate fails.

If the phrase appears only in output/DRAFT.md, where should the fix go?

Route the fix to the earliest responsible polisher: writer-selfloop, subsection-polisher, or draft-polisher depending on where the phrase occurs. The report will specify the recommended target.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational
post-merge-voice-gate skill by willoscar/research-units-pipeline-skills | VeilStrat