arkcli-datasets

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from local .jsonl files and remote TOS objects. Ingestion points: Data enters the agent context via local file mapping and remote object mapping, and through the --preview feature which reads content from objects (SKILL.md, references/commands.md). Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions embedded within dataset content. Capability inventory: The skill executes arkcli shell commands to perform file operations and network requests for dataset management and training job creation. Sanitization: No validation or filtering of external dataset content is mentioned.
  • [COMMAND_EXECUTION]: The skill facilitates the construction and execution of shell commands using the arkcli utility based on user-supplied parameters like dataset IDs and file paths.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 10:31 PM
Security Audit — agent-trust-hub — arkcli-datasets