stitch-mcp-upload-design-md
Pass
Audited by Gen Agent Trust Hub on Jul 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to use the
Bashtool to execute thebase64command on the local filesystem. While intended for encoding, the availability of a shell tool allows for broader command execution capabilities. - [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via the
DESIGN.mdfile it processes. - Ingestion points: Local file
DESIGN.mdis read and its content is processed for Base64 encoding and API upload. - Boundary markers: There are no boundary markers or instructions to treat the file content as data only, potentially allowing embedded instructions to be executed by the agent.
- Capability inventory: The skill possesses
Bashaccess and the ability to interact with thestitchAPI tools, which could be abused if the agent follows malicious instructions in the design file. - Sanitization: The skill does not implement any sanitization or validation of the markdown content before processing.
Audit Metadata