The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions recommend using system utilities like security (macOS) and secret-tool (Linux) to retrieve 2FA secrets from the host's keychain or secret store. This is a secure and intended workflow to provide input to the TOTP generator without exposing secrets in command-line arguments.
[EXTERNAL_DOWNLOADS]: The bundled totp.py script features a clock synchronization diagnostic (--check-clock) that performs a network request to https://www.cloudflare.com by default. Cloudflare is a well-known service, and the request is used only to retrieve the HTTP Date header for comparison with the local clock.
[DATA_EXFILTRATION]: While the skill manages highly sensitive TOTP shared secrets, it includes explicit design constraints to prevent leakage. It mandates that secrets be passed via stdin or files instead of command-line arguments and strictly forbids the agent from including the raw secret in its final response.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data, such as otpauth:// URIs and secret files.
Ingestion points: Secret strings and URIs read from files, environment variables, or stdin within scripts/totp.py.
Boundary markers: The script does not use specific boundary markers in the prompt, but it enforces strict parsing of the incoming data.
Capability inventory: The skill has the capability to read local files, execute system commands (keychain retrieval), and perform diagnostic network HEAD requests.
Sanitization: The script performs validation on all inputs, including casting URI parameters to integers and restricting HMAC algorithms to a specific allowed list (SHA1, SHA256, SHA512).

totp-generate