Adds unslop to the Communication & Writing section in alphabetical order (after "Twitter Algorithm Optimizer").

What it does: Claude Code plugin + CLI that humanizes AI output by stripping the stylometric residue from RLHF — sycophancy openers, stock vocab, hedging stacks, tricolon padding, em-dash pileups — while preserving code blocks, URLs, headings, and YAML frontmatter byte-exact. Six intensity modes (subtle / balanced / full / voice-match / anti-detector) toggle with /unslop.

Why it fits this list: It's a real Claude Code skill (installable via /plugin marketplace add MohamedAbdallah-14/unslop + /plugin install unslop), not a general-purpose tool masquerading as one. The skill injects activation rules into Claude's context at SessionStart, re-emits reinforcement banners at turns 8/16/24 to beat persona drift (per RMTBench/HorizonBench findings), and tracks the active mode in ~/.claude/.unslop-mode.

Measured results: Blind LLM-judge humanness preference 21/21 (100%) on a 7-fixture suite, 3 independent runs per fixture, randomized A/B sides. 89.1% AI-ism reduction vs baseline. Reproducible with python3 evals/perceived_humanness.py --runs 3.

Use case: Cover letters, personal statements, LinkedIn posts, blog posts, marketing copy — anywhere a human reader judges you on how the text sounds. Explicitly NOT for code, legal text, medical advice, or runbooks (see the "When it actually matters" section in the README).

MIT-licensed. Six-assistant support via a single synced plugin (Claude Code, Cursor, Windsurf, Cline, Gemini CLI, OpenAI Codex).

Repo: https://github.com/MohamedAbdallah-14/unslop