excel-pivot-wizard
Fail
Audited by Gen Agent Trust Hub on Feb 15, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- COMMAND_EXECUTION (HIGH): The
SKILL.mdfile definesBash(cmd:*)in theallowed-toolssection, granting the agent unrestricted shell access. This allows execution of any system command and violates the principle of least privilege. \n- REMOTE_CODE_EXECUTION (MEDIUM): Thescripts/README.mdfile describes a process for creating and executing local scripts ('chmod +x' and execution 'during skill activation'), which could be used to run arbitrary code. \n- PROMPT_INJECTION (HIGH): The skill possesses a high-risk surface for indirect prompt injection. \n - Ingestion points: Tabular data and spreadsheets as specified in
SKILL.mdinstructions. \n - Boundary markers: None; there are no instructions to the agent to treat external data as untrusted or to use delimiters. \n
- Capability inventory:
Bash(cmd:*),Write,Edit,Grep,Globtools are available to the agent. \n - Sanitization: Absent; the skill lacks any mention of input validation, escaping, or filtering for data processed from spreadsheets.
Recommendations
- AI detected serious security threats
Audit Metadata