Today's Agent Skill: Automated Expense Report Audit
What It Does
Finance teams burn hours each week manually checking expense reports for policy violations, duplicate receipts, and missing documentation — and most violations still slip through. An AI agent can flag every anomaly before a report reaches a human approver, cutting review time and catching the mistakes people miss.
How It Works
The skill ingests a batch of submitted expense line items and receipts, cross-references each against the company's expense policy (spend caps, category rules, required documentation) and prior submissions, then flags duplicates, out-of-policy charges, and missing receipts with a confidence score and a one-line reason for each flag.
How to Deploy It
Drop the SKILL.md file into any SKILL.md-compatible agent's skills directory and point its Input section at your expense policy doc and the export path for submitted reports (CSV, PDF, or API pull from your expense tool). Trigger it manually before each approval cycle, or wire it to a cron/webhook so it runs automatically whenever a new batch of reports is submitted.
SKILL.md — Ready to Deploy
## Description
Audits a batch of submitted expense reports against company policy, flags duplicate submissions, out-of-policy spend, and missing documentation, and produces a prioritized exceptions list for a human approver — before reports reach the approval queue.
## Trigger
Runs when a new batch of expense reports is submitted, on a scheduled cadence (e.g., daily or weekly before the approval cutoff), or on manual invocation by a finance reviewer.
## Input
- Expense policy document (spend caps by category, receipt requirements, approved vendors)
- Current batch of expense reports (line items, amounts, categories, attached receipts)
- Prior 90 days of approved expense history (for duplicate detection)
## Steps
1. Parse each expense report into line items with amount, category, date, vendor, and receipt status.
2. Check each line item against the policy document for category caps, approved vendor lists, and required documentation.
3. Cross-reference each line item against prior submissions to detect duplicate claims (same vendor, amount, and date within a tolerance window).
4. Flag missing or illegible receipts above the policy's documentation threshold.
5. Score each flag by sCopy the full SKILL.md and drop it into your agent's skills directory to activate this skill.