maintainx-local-dev-loop

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill provides templates for an integration that processes data from the MaintainX API (e.g., work order titles, descriptions, and metadata). Malicious data stored in the SaaS platform could attempt to influence the agent's behavior during interactive debugging or test execution if the agent processes the tool output.\n
  • Ingestion points: External API responses are retrieved and displayed in src/client.ts, tests/api.test.ts, and scripts/repl.ts.\n
  • Boundary markers: The provided code templates do not include delimiters or specific instructions for the agent to ignore potentially malicious content within the API data.\n
  • Capability inventory: The skill uses Bash(npm:*), Bash(node:*), and Write tools, providing a path for system interaction if an injection is successful.\n
  • Sanitization: There is no evidence of sanitization or validation of the external API content before it is processed or printed to the console.\n- [DYNAMIC_EXECUTION]: The skill implements an interactive REPL and uses TypeScript execution tools that allow for dynamic code evaluation in the local environment.\n
  • The scripts/repl.ts file utilizes the node:repl module to create an interactive shell where the MaintainXClient is exposed, allowing arbitrary JavaScript execution.\n
  • package.json scripts use tsx and ts-node to execute TypeScript files directly, which is standard for development loops but constitutes dynamic execution of generated code.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 03:45 AM
Security Audit — agent-trust-hub — maintainx-local-dev-loop