maple-onboarding-style
Pass
Audited by Gen Agent Trust Hub on Oct 19, 2026
Risk Level: SAFECREDENTIALS_UNSAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The instructions recommend hardcoding the
MAPLE_KEYandMAPLE_ENDPOINTdirectly in the bootstrap source code instead of using environment variables or secret management tools. While the key is described as a 'public ingest key', hardcoding credentials increases the risk of accidental exposure via version control systems. - [DATA_EXFILTRATION]: The skill configures OpenTelemetry to send traces, metrics, and logs to an external vendor endpoint (
https://ingest.maple.dev). This telemetry data can include sensitive business information such as tenant IDs, order IDs, and potentially user-generated content from LLM interactions. - [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for capturing LLM inputs and outputs for telemetry. While it notes 'hideInputs' and 'hideOutputs' as safe defaults, the ability to transmit full prompts and completions to an external service creates a data exposure surface for any potentially malicious content processed by the LLM.
Audit Metadata