skills/openclaw/skills/runpod/Gen Agent Trust Hub

runpod

Fail

Audited by Gen Agent Trust Hub on Feb 15, 2026

Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [CREDENTIALS_UNSAFE] (HIGH): The skill accesses sensitive private SSH keys located at ~/.ssh/runpod_key and ~/.runpod/ssh/RunPod-Key.
  • [COMMAND_EXECUTION] (HIGH): The mount_pod.sh script executes shell commands (mkdir, sshfs, runpodctl) using a user-supplied pod_id. The lack of input validation allows for potential path traversal (e.g., using ../ in a pod ID) to affect the host filesystem.
  • [EXTERNAL_DOWNLOADS] (MEDIUM): Requires installation of the runpodctl binary from a non-trusted third-party Homebrew tap (runpod/runpodctl).
  • [Indirect Prompt Injection] (HIGH): Vulnerability surface identified in scripts/mount_pod.sh where untrusted user input is directly used in privileged operations. 1. Ingestion points: pod_id argument. 2. Boundary markers: Absent. 3. Capability inventory: mkdir, sshfs, runpodctl, mount. 4. Sanitization: Absent; no validation is performed on the pod_id variable before it is used to construct file paths and command arguments.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 15, 2026, 03:25 PM