plugin.json

{
  "name": "writing-quality",
  "version": "1.1.0",
  "description": "AI writing pattern detection and 5-dimension quality scoring with content-type-aware thresholds. Deterministic checks for banned phrases/structures + LLM scoring for voice, rhythm, authenticity. Landing pages: 28/50 threshold with adjusted Rhythm/Trust/Authenticity. Prose: 35/50. Sources: humanizer (29 patterns) + stop-slop (5-dim scoring). Calibrated against Stripe, Linear, Basecamp, Vercel, Notion.",
  "tools": [
    {
      "name": "check_writing_quality",
      "description": "Score content on 5 dimensions (Directness, Rhythm, Trust, Authenticity, Density) and detect AI writing patterns. Content-type-aware: landing pages use 28/50 threshold with adjusted Rhythm/Trust/Authenticity scoring; prose uses 35/50. Deterministic pattern matching for banned phrases/structures, LLM for judgment-based scoring.",
      "inputSchema": {
        "type": "object",
        "properties": {
          "content": {
            "type": "string",
            "description": "The text content to audit"
          },
          "context": {
            "type": "string",
            "description": "Optional brand context (voice rules, audience, constraints)"
          },
          "mode": {
            "type": "string",
            "enum": ["full", "patterns-only", "score-only"],
            "description": "Audit mode: full (patterns + scoring), patterns-only (deterministic only, no LLM), score-only (LLM scoring only). Default: full."
          }
        },
        "required": ["content"]
      },
      "handler": "node handler.js"
    }
  ]
}