import-art
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use the
Bashtool to executemkdir -pandcpcommands. These commands incorporate variables like{source_file}and{album}which are derived from the user-provided$ARGUMENTS. - Evidence: SKILL.md contains shell command templates:
mkdir -p {audio_root}/artists/{artist}/albums/{genre}/{album}andcp "{source_file}" "...". - [INDIRECT_PROMPT_INJECTION]: The skill provides a surface for indirect injection by ingesting untrusted user input via
$ARGUMENTSand using it to drive shell command generation without explicit boundary markers or sanitization requirements. - Ingestion points:
$ARGUMENTS(containing<file-path>and<album-name>) in SKILL.md. - Boundary markers: Absent.
- Capability inventory:
Bashtool usage for directory creation and file copying in SKILL.md. - Sanitization: Absent.
Audit Metadata