xbert-policy-review

{
  "name": "xbert-policy-review",
  "displayName": "XBert Policy Review",
  "description": "Weekly review of posted spend against policy XBerts on supported categories (out-of-policy account, banned vendor, missing receipt, duplicate transaction, receipt forensic flag).",
  "author": {
    "name": "XBert Intelligence",
    "email": "[email protected]"
  },
  "homepage": "https://intelligence.xbert.io/plugins/xbert-policy-review",
  "license": "MIT",
  "keywords": [
    "policy",
    "spend",
    "compliance",
    "forensic",
    "receipts",
    "review",
    "xbert"
  ],
  "x-xbert": {
    "displayName": "XBert Policy Review",
    "tagline": "Weekly review of posted spend against policy XBerts on supported categories — out-of-policy account, banned vendor, missing receipt, duplicate transaction, receipt forensic flag.",
    "audience": [
      "bookkeepers",
      "accountants"
    ],
    "categories": [
      "review",
      "policy",
      "spend",
      "compliance"
    ],
    "prerequisites": [
      "XBert account",
      "Client connected to a ledger",
      "Policy XBerts linked to the Policy Review agent in the Connect portal"
    ],
    "screenshots": [],
    "demoVideo": null,
    "includes": {
      "skills": 1,
      "commands": 1,
      "mcpTools": 0
    },
    "longDescription": "Spend policy is enforced deterministically by Custom XBerts on the data XBert MCP exposes today — out-of-policy account postings, banned-vendor postings, missing-receipt flags, duplicate transactions (same vendor + amount + date), and the receipt forensic signals on captured images (copy-move, metadata tampering, greyscale, image-of-image). The XBerts do the detection; this plugin does not.\n\nXBert Policy Review orchestrates the weekly periodic review of what the policy XBerts flagged. Firings are grouped by category, sub-grouped by vendor or account where it tightens the story, and each firing carries a recommended next action (recode, request supporting doc, accept-as-correct, escalate). The user writes the communication; the plugin does the prep and the audit pack.\n\nOut of scope (no backing data in MCP today): per-employee patterns, per-cardholder breaches, after-hours posting attribution, drafted spender follow-up copy, manual journal forensics, expense-claim review, mileage-over-threshold per employee. If a Custom XBert lands later that captures any of these signals, the plugin will surface them; until then the agent says so honestly rather than invent the signal.",
    "useCases": [
      {
        "title": "Weekly spend review",
        "description": "Every Monday morning, sweep every policy XBert that fired across the prior week and walk the recommended action per firing."
      },
      {
        "title": "Pre-period-close policy sweep",
        "description": "Before the month closes, ensure every policy violation is either resolved with the user or escalated for write-off — per firing."
      },
      {
        "title": "Vendor-anomaly investigation",
        "description": "When the same vendor keeps firing across the period, the per-vendor recurrence count surfaces the pattern and drives a vendor-master decision."
      },
      {
        "title": "Receipt forensic-flag review",
        "description": "Capture XBerts catch copy-move, metadata-tampering, greyscale and image-of-image signals on captured receipts; this plugin surfaces them and proposes the next action."
      }
    ],
    "benefits": [
      {
        "icon": "ShieldCheck",
        "title": "Policy XBerts enforce deterministically",
        "description": "Out-of-policy account, banned vendor, missing receipt, duplicate transaction, receipt forensic — every rule is owned by a deterministic XBert. The plugin reasons only over what fired."
      },
      {
        "icon": "Search",
        "title": "Forensic flags surfaced specifically",
        "description": "Receipt forensic signals (copy-move, metadata tampering, greyscale, image-of-image) caught by Capture XBerts are named explicitly per receipt — not 'looks suspicious'."
      },
      {
        "icon": "FileText",
        "title": "Audit pack output",
        "description": "Outputs a workpaper grouped by violation category with per-firing recommended actions and a QMS block — filable straight into the client folder."
      }
    ],
    "workflow": [
      "Type /policy-review in Claude Code or Claude Desktop",
      "Confirm the client and the review period (default: prior week)",
      "Plugin resolves the linked policy XBerts for the period — if none are linked, the review stops and the user is prompted to configure linked XBerts first",
      "Firings are grouped by supported category (with vendor or account sub-grouping where useful) and each firing carries a recommended next action",
      "Build the working-paper payload, render via xbert-working-paper:render-docx, and report the saved path"
    ]
  }
}