pp-walkingpad

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install a CLI tool from the author's official sources.\n
  • Fetches from the @mvanhorn/printing-press-library package on NPM using npx.\n
  • Offers a fallback installation from the github.com/mvanhorn/printing-press-library repository using go install.\n- [COMMAND_EXECUTION]: The agent is instructed to use the walkingpad-pp-cli binary to perform various tasks.\n
  • Includes commands for physical device control such as run, stop, and set-speed.\n
  • Implementation includes safety flags like --confirm-physical-effect to prevent accidental actuation.\n- [PROMPT_INJECTION]: The skill creates an indirect prompt injection surface by ingesting data from external hardware and local history files.\n
  • Ingestion points: Data retrieved from the treadmill via Bluetooth LE and stored in local JSON files (e.g., daily.json).\n
  • Boundary markers: Instructions do not specify delimiters to separate untrusted device data from agent instructions.\n
  • Capability inventory: The skill can execute subprocesses that control physical hardware (run, stop, set-speed).\n
  • Sanitization: No specific sanitization or validation of the treadmill-provided data is mentioned in the agent instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 05:47 PM
Security Audit — agent-trust-hub — pp-walkingpad