qwenpaw-usage
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONPRIVILEGE_ESCALATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents the 'Heartbeat' feature which reads content from
~/.qwenpaw/HEARTBEAT.mdand processes it as an AI prompt on a scheduled basis. - Ingestion points: The file
~/.qwenpaw/HEARTBEAT.mdacts as a stored prompt that is ingested by the agent context at regular intervals. - Boundary markers: There are no delimited structures or explicit 'ignore embedded instructions' warnings specified for the content of the heartbeat file.
- Capability inventory: The skill utilizes shell command execution (via the
qwenpawCLI and Docker), file writing, and network status checks. - Sanitization: No sanitization or validation logic is described for the content stored in the heartbeat file.
- [COMMAND_EXECUTION]: The skill provides numerous shell commands for the agent to manage background processes, schedule tasks, and control Docker containers.
- Evidence: Instructions include the use of
qwenpaw daemon reload-config,qwenpaw cron create, anddocker compose up -dto manage the QwenPaw environment. - [PRIVILEGE_ESCALATION]: The skill contains a command that utilizes administrative privileges to perform file operations within protected system paths.
- Evidence: The deployment instructions suggest using
sudo tar xzf qwenpaw-workspace.tar.gz -C "$MOUNT_PATH"to restore data into a Docker volume mount point.
Audit Metadata