wrap
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [SAFE] (SAFE): No malicious patterns, data exfiltration, or unauthorized access vectors were detected within the skill's logic or references.
- [COMMAND_EXECUTION] (LOW): The skill executes standard git commands (
git status,git diff) to understand the current workspace state. These are benign, read-only operations in this context. - [PROMPT_INJECTION] (LOW): The skill ingests untrusted data from the workspace (via git output) and user notes into its sub-agents. Although explicit boundary markers and sanitization are not defined in the prompts, the risk of indirect prompt injection is mitigated by a multi-agent validation phase and a mandatory human-in-the-loop confirmation step (
AskUserQuestion) before any significant actions (commits, file modifications) are executed.
Audit Metadata