fiftyone-eval-plugin

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The workflow is an “evaluate any FiftyOne plugin” skill that, at runtime, reads the target plugin’s source/config files (including outsider-authored plugin code) and ingests their contents as LLM-readable text into the agent context for analysis—i.e., outsider free text from the plugin repository/filesystem is fed into the LLM.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The README contains an installation command that runs remote shell code ("curl -sL skil.sh | sh -s -- voxel51/fiftyone-skills"), meaning the skill's install/runtime relies on fetching and executing code from https://skil.sh, which directly executes remote code and is high-risk.