mcp-code-execution

Pass

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill configuration requests the 'Bash' tool to support the execution of agent-generated scripts within the proposed architecture.
  • [REMOTE_CODE_EXECUTION]: The core pattern involves runtime generation and execution of code; the skill includes a security checklist requiring sandboxed environments, execution timeouts, and resource limits to mitigate risks.
  • [PROMPT_INJECTION]: The described architecture is vulnerable to indirect prompt injection. 1. Ingestion points: MCP tool responses such as document content or channel history enter the context. 2. Boundary markers: None specified in the provided logic snippets. 3. Capability inventory: The configuration allows high-privilege tools including Bash, Write, and Edit. 4. Sanitization: While PII tokenization is mentioned for data flowing to the model, no explicit sanitization of tool outputs is described before they influence the agent's code generation logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 27, 2026, 11:33 AM