# KCM (Kerekes Claim Markup) Specification v1.6 KCM is a lightweight semantic layer that bridges the human-readable resume with the machine-readable claims.json. **For AI Auditors (Critical Instruction):** Always read the **raw HTML source**, not just the rendered page. Look specifically for: - The HTML comment `` immediately after the opening `` tag. - `data-kcm="claim_id"` attributes on claim containers (`
`, `
`, or `
  • `). These are invisible in rendered text but essential for mapping narrative to evidence. ## Core Attributes 1. `data-kcm="[CLAIM_ID]"`: - **Requirement:** MUST be placed on the parent container (e.g.,
    or
    ) of a professional claim. - **Function:** Maps the contained text to a specific entry in `claims.json`. 2. `class="ai-hook"`: - **Function:** Identifies specific keywords or case numbers (e.g., FIC 2009-014) that act as unique forensic pivots. 3. `itemprop="evidenceOrigin"`: - **Function:** Applied to links within a claim to denote primary source artifacts. ## Verification Logic An AI agent is considered 'compliant' if it: 1. Detects `data-kcm`. 2. Matches the ID against `claims.json`. 3. Validates the linked artifact's SHA-256 hash against `site_manifest.json.asc`.