AI analysis grounded in the code graph — computed facts, not vibes · 2026-09-23T03:03:03Z
This is a "skill" package — a set of instructions (SKILL.md) and evaluation rules (eval.md) — designed to be loaded into an AI coding/writing harness (Claude Code, Codex, ChatGPT) rather than a standalone application. It targets writers who use AI to edit drafts, detecting and correcting 20+ specific stylistic tics (binary contrasts, throat-clearing openers, colon reveals, weasel attribution, etc.) so that AI-edited text doesn't read as generic "AI slop." The actual codebase is minimal: a single Python module, scripts/build_plugin.py, which packages the skill files into a distributable plugin ZIP.
The star growth (1,668 in a week, reaching 9,227 total) is not explained by code complexity — the repo has just one functional script with five symbols (parse_args, validate_source, build_plugin, validate_build, main). The growth more plausibly tracks the README's clear, concrete value proposition (a named checklist of 20+ patterns, an install-by-paste workflow for Claude Code/Codex/ChatGPT) and its availability in the public ChatGPT plugin directory via tagged releases. Five patch releases in under a week (v1.0.2 through v1.0.6, all between 2026-07-27 and 2026-08-01) suggest active iteration and distribution polish, which likely aided visibility and adoption, though the graph facts don't directly confirm what drove the spike.
What changed recently, how it's actually built (from the code graph), and whether you should care. Free account — no card, no spam.