nimble-agent-builder

Pass

Audited by Gen Agent Trust Hub on Apr 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses Bash to perform project environment checks and dynamically generates Python scripts for site investigation and agent validation as described in SKILL.md and the agent lifecycle documentation.
  • [REMOTE_CODE_EXECUTION]: Configuration instructions for Cursor and VS Code involve executing a remote MCP server via npx mcp-remote which connects to a vendor-controlled API endpoint at https://mcp.nimbleway.com/mcp.
  • [EXTERNAL_DOWNLOADS]: The skill setup and reference documentation direct the installation of official vendor SDKs and CLI tools from standard registries (NPM, PyPI, GitHub).
  • [PROMPT_INJECTION]: The site investigation process ingests content from external web pages via Playwright and the Nimble CLI, creating a potential surface for indirect prompt injection attacks.
  • Ingestion points: Playwright page.goto and nimble extract commands in SKILL.md investigation workflows.
  • Boundary markers: Absent from the investigation reports and tool output processing.
  • Capability inventory: Local script execution, subprocess calls via Bash, and network operations via the Nimble platform.
  • Sanitization: No specific filtering or escaping is applied to web-ingested content before processing by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 21, 2026, 06:37 PM