linux-file-integrity

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONPERSISTENCEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous bash commands for system administration, including package management (apt install), security tool initialization (aideinit), and log management. These are within the scope of a file integrity monitoring tool.\n- [PRIVILEGE_ESCALATION]: The instructions make extensive use of sudo to perform administrative tasks. While this grants elevated permissions, it is necessary for the stated purpose of configuring system-wide integrity monitoring.\n- [PERSISTENCE]: The skill configures scheduled tasks using cron.daily and systemd timers (aide-check.timer) to ensure continuous monitoring. This is a functional requirement for the tool's purpose.\n- [EXTERNAL_DOWNLOADS]: The skill installs software using standard Linux package managers (apt, dnf). These rely on official and trusted operating system repositories.\n- [INDIRECT_PROMPT_INJECTION]: The skill involves reading and triaging drift reports that include filenames and file attributes. This represents a potential surface where specially crafted filenames could attempt to influence agent behavior.\n
  • Ingestion points: AIDE drift reports (aide --check output) containing file paths and attributes are processed by the agent in SKILL.md and references/aide-reference.md.\n
  • Boundary markers: The skill provides instructions for manual triage and comparison against trusted package manifests, acting as a procedural boundary.\n
  • Capability inventory: The skill uses subprocess calls for aide, apt, dnf, and file-write operations for configuration and logging.\n
  • Sanitization: No explicit programmatic sanitization of filenames is mentioned in the provided scripts or instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 02:40 AM