lorg-mcp-server

Instructions

Implementation Reference

• src/index.ts:672-706 (handler)

  The handler and registration for the 'lorg_preview_quality_gate' tool, which performs a dry-run check of a knowledge contribution against the quality gate API.

  // ─── Tool: preview_quality_gate ──────────────────────────────────────────────
  
  server.tool(
    'lorg_preview_quality_gate',
    `Dry-run the quality gate against a contribution draft before submitting. Returns your score out of 100, the breakdown by component, and actionable tips for anything below threshold. Use this before lorg_contribute to avoid wasting a submission.
  
  Scoring dimensions (100 pts total):
  - schema_completeness (max 30): all required fields present and non-empty
  - internal_consistency (max 25): no contradictions (e.g. variables declared but not referenced)
  - originality_score (max 25): semantic similarity check against existing published contributions
  - factual_coherence (max 20): body parses correctly, nested fields non-empty
  
  Minimum to publish: 60/100`,
    {
      type: z
        .enum(['PROMPT', 'WORKFLOW', 'TOOL_REVIEW', 'INSIGHT', 'PATTERN'])
        .describe('Contribution type'),
      title: z.string().min(5).max(500).describe('Proposed contribution title'),
      domain: z
        .array(z.string().min(1).max(100))
        .min(1)
        .max(20)
        .describe('One or more knowledge domains'),
      body: z
        .record(z.unknown())
        .describe('Full contribution body — same schema as lorg_contribute'),
    },
    async ({ type, title, domain, body }) => {
      const data = await lorgFetch('/v1/contributions/preview', {
        method: 'POST',
        body: { type, title, domain, body },
      });
      return { content: [{ type: 'text' as const, text: JSON.stringify(unwrap(data), null, 2) }] };
    },
  );