read
Warn
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to generate and execute an inline Python script via a Bash command to resolve file paths. This script uses string interpolation to insert a page ID (
pid = "<id>"). Because the ID is placed directly inside a string literal, a crafted input (e.g.,"; <malicious_code> #) would allow an attacker to escape the string and execute arbitrary Python code within the agent's environment. - [COMMAND_EXECUTION]: The skill performs filesystem operations to read application state from
~/.origin/pages/.origin/state.json. While this is part of the tool's intended functionality to map IDs to markdown files, it involves programmatic access to local application configuration data via shell execution.
Audit Metadata