sardis-cards

What This Skill Does

The Sardis Cards skill provides a secure, programmatic interface for AI agents to manage virtual payment cards. Designed specifically for autonomous spending, it enables agents to execute real-world transactions while adhering to strict financial governance protocols. By integrating with the Sardis API, the skill allows for the creation of virtual cards, granular spending limits, and automated security controls. It acts as a financial bridge, allowing agents to pay for SaaS subscriptions, cloud infrastructure, or other third-party services without exposing a primary corporate account. The system emphasizes security, requiring automatic card freezing upon detection of anomalies and enforcing strict spending caps for every individual agent.

Installation

To integrate this skill into your environment, use the OpenClaw CLI tool to pull from the repository:

clawhub install openclaw/skills/skills/efedurmaz16/sardis-cards

Once installed, ensure your SARDIS_API_KEY is exported as an environment variable in your agent runtime: export SARDIS_API_KEY=sk_your_key_here.

Use Cases

• Cloud Infrastructure Auto-Scaling: Allow an agent to spin up and pay for temporary cloud GPU instances for heavy compute tasks with a pre-set daily budget.
• SaaS Subscription Management: Automate the procurement of software seats for team members, with usage-based spending limits.
• Vendor Purchasing: Assign unique virtual cards to specific vendors to isolate spending data and prevent unauthorized merchant charges.
• Fraud Prevention: Implement a 'fail-closed' mechanism where an agent automatically freezes its own spending capabilities if an unexpected billing pattern is detected.

Example Prompts

1. "Create a new virtual card for cloud services with a $500 monthly limit and assign it to the infrastructure agent."
2. "Check the current spending on card card_abc123 and freeze it if it exceeds 80% of the monthly limit."
3. "List all recent transactions for the SaaS subscription card and unfreeze it if no suspicious activity is detected."

Tips & Limitations

• Security First: Always treat the full card number response as highly sensitive data. Do not log the output of the /reveal endpoint.
• Spending Limits: Always define explicit per_transaction, daily, and monthly limits. Never operate without these constraints.
• Merchant Controls: Utilize allowed_categories to restrict the card usage to only the intended business operations, reducing the attack surface.
• Monitoring: Regularly poll the card status and current spend to ensure agent activity remains within expected business parameters.