start-new-sdk-project

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [Indirect Prompt Injection] (HIGH): The skill is vulnerable to indirect prompt injection through its processing of external OpenAPI specifications.\n
  • Ingestion points: The --schema (-s) flag accepts untrusted input from arbitrary URLs and local file paths.\n
  • Boundary markers: None present to distinguish between trusted instructions and untrusted data within the processed schema.\n
  • Capability inventory: The speakeasy quickstart command performs filesystem write operations (creating .speakeasy/workflow.yaml and the SDK output directory) and generates executable code.\n
  • Sanitization: No sanitization or validation of the schema content is performed by the skill before it is passed to the generation tool.\n- [Command Execution] (MEDIUM): The skill executes the speakeasy CLI with multiple parameters derived from user input. While necessary for its function, this provides a surface for argument manipulation if the agent does not strictly validate the inputs.\n- [Credentials Unsafe] (INFO): Documentation references the use of SPEAKEASY_API_KEY. While no secrets are hardcoded, the skill relies on the presence of sensitive credentials in the environment.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 02:28 AM