linux-access-control
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides a wide range of administrative commands for executing system-level changes, including user management (adduser, usermod, deluser), group management, and file permission resets (chmod, chown). These are intended for use by a system administrator or an agent with equivalent privileges to perform standard host maintenance.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests data from local system files, creating a surface for indirect prompt injection if those files contain attacker-controlled content.\n
- Ingestion points: The skill reads system files such as /etc/passwd, /etc/group, and authorized_keys across various user home directories (SKILL.md, references/users-sudoers-pam.md).\n
- Boundary markers: Absent. There are no explicit instructions or delimiters provided to the agent to help distinguish between administrative metadata and potential instructions embedded within fields like SSH key comments or user GECOS data.\n
- Capability inventory: The skill possesses extensive capabilities to modify system state, including user creation, password modification, and process termination (pkill).\n
- Sanitization: The skill does not provide mechanisms to sanitize or filter the content of system files before they are processed by the agent context.
Audit Metadata