agent-notion

Warn

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill includes authentication commands (agent-notion auth import-desktop and agent-notion auth import-browser) that extract sensitive Notion session credentials (token_v2) directly from the application data directories and browser cookie stores on the local filesystem. This includes access to profiles from browsers such as Chrome, Firefox, Safari, and Edge.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from Notion pages, databases, and comments which may contain malicious instructions designed to influence the agent's behavior.
  • Ingestion points: SKILL.md (commands: page get, block list, database query, comment list).
  • Boundary markers: Absent. The instructions do not define specific delimiters or security warnings to separate untrusted Notion content from the system context.
  • Capability inventory: SKILL.md (commands: page update, page trash, page archive, block append, block update, block delete, block replace). These capabilities allow an attacker controlling Notion content to potentially manipulate the workspace or perform destructive actions.
  • Sanitization: Absent. There is no evidence of input validation or content filtering specified in the instructions.
  • [COMMAND_EXECUTION]: The skill relies on the execution of the agent-notion binary via the Bash tool to perform its operations.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 31, 2026, 09:40 PM
Security Audit — agent-trust-hub — agent-notion