conorbronsdon/avoid-ai-writing v3.34.0

on GitHub

Added

• Add the avoid-ai-writing command-line interface (package bin) for scoring one file or piped text as JSON, with --context and --source-mode options, --help, a -- end-of-options separator, and usage/I/O errors on stderr with exit code 2, covered by child-process tests (#158).
• Add a pattern proposal issue form so a new rule arrives with a should-fire
  example, a must-not-fire example, and its false-positive risk; link it from CONTRIBUTING.md.

Changed

• Add repository-local SSOT CI checks for the detector's Node requirement,
  advisory discovery, and drift controls; pin the existing promo checker.
• Point the bundled house-style examples at the canonical public README so the link still works when the skill package is installed without the repository root.
• Keep em-dash overuse as a P2 writing-quality flag while excluding it from the authorship score, label, probabilities, confidence, and classification (#73). Existing rate thresholds and carve-outs are unchanged. Scores may be lower for text where em-dash overuse previously contributed weight.
• Keep paired prose quotes around bare URLs visible to quote normalization even when the URL contains an unmatched opening parenthesis. Preserve internal URL apostrophes and explicit Markdown link destinations.
• Add a GitHub follow invitation to the README's maintainer section.