The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interacts with the Stripe CLI through a Python wrapper script (scripts/stripe_cli.py). The implementation is secure as it uses subprocess.run with list-based arguments and validates all user-supplied inputs (such as resource paths, subscription IDs, and clock IDs) against strict regular expression allowlists (READ_PATH_PATTERN, CLOCK_ID_PATTERN, SUBSCRIPTION_ID_PATTERN) before execution. This prevents the possibility of shell command injection.\n- [CREDENTIALS_UNSAFE]: The wrapper script accesses the Stripe CLI's local configuration file (~/.config/stripe/config.toml) or the STRIPE_API_KEY environment variable to authenticate requests. Security is maintained by a dedicated check_key function that explicitly identifies and rejects production keys (prefixes sk_live_ and rk_live_), ensuring the tool only operates in test mode.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes data from Stripe API responses, which may contain malicious instructions embedded in metadata or descriptions by external actors.\n
Ingestion points: Stripe API resource data retrieved via the stripe_cli.py wrapper script and processed in SKILL.md.\n
Boundary markers: The instructions explicitly direct the agent to treat Stripe content as untrusted data and to report any embedded instructions to the user as potential security concerns (CWE-1427) rather than obeying them.\n
Capability inventory: The skill allows reading various Stripe resources and advancing existing test clocks via a sanitized wrapper.\n
Sanitization: The skill utilizes strict regex validation for all identifiers and resource paths passed to the underlying CLI tool.

using-stripe-cli