Adds chatbot-deference phrases, stacked-negation parallelisms,
and AI vocabulary (meticulous, bolstered, vibrant, etc.).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Rewrite SKILL.md: remove ~80 lines of duplicated AI patterns (already in
taboo-phrases.md), duplicated rubric table (already in rubric.md), and
redundant examples. SKILL.md is now a workflow/decision layer that points
to references for details.
- Add progressive disclosure guidance (when to read each reference file)
- Simplify output format: minimal by default, detailed on --strict
- Expand description for better trigger coverage
- Add TOC to taboo-phrases.md (673-line file)
- Add evals/evals.json with 4 test cases covering LinkedIn post, technical
article with fact preservation, warm-preset email, and heavy-slop paragraph
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add em-dash detection to scanner (soft for single, hard for multiple per paragraph)
- Add ~40 new banned phrases: resonates with, speaks volumes, at the intersection of, etc.
- Add structural patterns: colon reveals, staccato fragmentation, So-openers, exclamation overuse
- Promote Punctuation Abuse to top-level section in taboo-phrases.md
- Fix contradictory examples that used em-dashes in "after" text
- Update SKILL.md workflow to explicitly require em-dash elimination in Pass 2
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Clayton Kim