pushary-cowork

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [DATA_EXFILTRATION]: The skill instructions specify sending task summaries and session metadata (agent name) to the Pushary service via the send_notification tool. This functionality is the primary purpose of the skill and targets the developer's official domain (pushary.com).
  • [INDIRECT_PROMPT_INJECTION]: The skill establishes a data ingestion surface where external input from a user's mobile device is returned to the agent's context via the wait_for_answer tool.
  • Ingestion points: Tool output from wait_for_answer which contains user-provided text from lock screen notifications.
  • Boundary markers: None identified in the skill instructions to distinguish external input from system instructions.
  • Capability inventory: The instructions indicate the agent has high-privilege capabilities including file modification, financial transactions, and external network communications.
  • Sanitization: The skill does not mention any validation or filtering of the content returned from the external notification service.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 07:27 PM
Security Audit — agent-trust-hub — pushary-cowork