agent-in-sync

Warn

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: MEDIUMPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill uses high-pressure language such as 'STOP' and 'NON-NEGOTIABLE' to override the agent's internal operational logic and force compliance with a specific external workflow.
  • [DATA_EXFILTRATION]: The tools search_before_fixing and submit_after_solving encourage the agent to send sensitive information, including source code, stack traces, and project identifiers like repository names, to an external database (agentinsync.com).
  • [PROMPT_INJECTION]: By requiring the agent to use and apply solutions from a collaborative, user-contributed knowledge base, the skill introduces a vulnerability to indirect prompt injection where untrusted content could influence agent actions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 10, 2026, 01:15 PM