The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill implements strong security boundaries for sensitive data. It explicitly requires using environment variables (e.g., MAKE_APP_KEY, MAKE_API_TOKEN) instead of hardcoding secrets. It mandates the redaction of Authorization headers, cookies, and tokens in all boundary logs. Furthermore, it prevents the exposure of raw backend file URLs and bearer tokens to the frontend by requiring a Service-side download proxy that validates sessions before attaching credentials.
[COMMAND_EXECUTION]: The instructions explicitly forbid calling CLI tools like makecli during service runtime in production environments, preventing potential security risks associated with shell execution in containers.
[INDIRECT_PROMPT_INJECTION]: The skill defines patterns for proxying Make AI assistant traffic. It includes safety-conscious error handling that converts malformed or non-public upstream errors into safe service errors, preventing the leakage of raw backend diagnostics to users.
[SAFE]: The skill architecture promotes robust request handling using AbortSignal propagation, ensuring that client disconnects result in the cancellation of downstream work, which prevents resource exhaustion attacks. It also addresses data integrity by requiring lossless JSON decoding for large integer IDs, preventing precision loss in sensitive record identifiers.

make-app-service