The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads and executes the @stripe/link-cli package from the npm registry using npx. This is the primary mechanism for all provided payment and wallet management functionality.
[COMMAND_EXECUTION]: Extensive use of shell commands to interact with the Stripe Link wallet, including authentication, payment method retrieval, and transaction execution.
[CREDENTIALS_UNSAFE]: The skill manages sensitive authentication tokens and session data stored locally at ~/.config/link-cli-nodejs/config.json. It also explicitly allows the agent to retrieve full payment card details, including PAN, CVC, and expiration dates, which are necessary for completing purchases.
[DATA_EXFILTRATION]: The mpp pay and spend-request create commands perform network operations to external merchant and API endpoints. This is the intended behavior for processing Machine Payments Protocol (MPP) transactions and submitting spend requests.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface through the mpp decode command, which processes external WWW-Authenticate header challenges.
Ingestion points: The --challenge argument in the mpp decode command accepts untrusted data from external web headers.
Boundary markers: None present to isolate the challenge string from the command logic.
Capability inventory: The skill has the capability to execute network requests via mpp pay and initiate financial transactions via spend-request create.
Sanitization: There is no explicit sanitization or validation mentioned for the challenge data before it is processed by the CLI tool.

link-cli