skills/lukastk/boxyard/boxyard-cli/Gen Agent Trust Hub

boxyard-cli

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill enables management of local and remote folders through the boxyard CLI, involving shell commands for syncing, deleting, moving, and cloning data.
  • [INDIRECT_PROMPT_INJECTION]: The tool processes configuration and metadata files (boxmeta.toml, conf/sync.toml, conf/.rclone_exclude) located within managed directories. An attacker who can control these files could potentially attempt to influence the agent's context.
  • Ingestion points: boxmeta.toml, conf/sync.toml, and conf/.rclone_exclude files as described in SKILL.md.
  • Boundary markers: Absent; the tool reads these files to determine synchronization behavior.
  • Capability inventory: Substantial file system and remote synchronization commands including sync, delete, rename, and force-push.
  • Sanitization: Not explicitly mentioned in the skill instructions.
  • [CREDENTIALS_UNSAFE]: The skill mentions sensitive configuration files (~/.config/boxyard/boxyard_rclone.conf) and environment variables (BOXYARD_RESTIC_PASSWORD) used for storage credentials and encryption, which are necessary for the tool's documented operation.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 04:51 PM