fiftyone-code-style
Fail
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The README.md file recommends an installation method using
curl -sL skil.sh | sh. This is a highly insecure pattern that downloads and executes shell scripts from a remote server without verification or integrity checks. - [EXTERNAL_DOWNLOADS]: The skill's installation instructions rely on fetching content from 'skil.sh', a domain that is not identified as a trusted vendor resource or a well-known service.
- [DYNAMIC_EXECUTION]: The templates provided in SKILL.md promote the use of
fou.ensure_package()andfou.lazy_import(). These functions allow for the dynamic installation of Python packages (e.g., 'open3d', 'pycocotools') at runtime, which bypassing static dependency management and increases the risk of supply chain attacks. - [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to search for code implementations in external repositories, such as GitHub, under the 'Avoid Redundant Code' section in SKILL.md. This introduces a surface for indirect prompt injection if the agent ingests malicious code from these external sources. Evidence Chain: (1) Ingestion Point: Remote search of GitHub repositories; (2) Boundary markers: Absent; (3) Capability inventory: Skill documentation describes tools for package installation and temporary directory creation; (4) Sanitization: Absent.
Recommendations
- AI detected serious security threats
Audit Metadata