The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to interpolate user-provided strings, such as model names and hub identifiers, directly into shell command templates. If the agent fails to properly escape these inputs, a malicious user could potentially execute arbitrary commands on the host system.
Evidence: found in SKILL.md Step 4: curl -s "http://localhost:8200/api/v1/models/jobs?model_name=<model-name>".
[REMOTE_CODE_EXECUTION]: The skill facilitates the download and immediate execution of a shell script from the author's public GitHub repository. While this is a vendor resource, the pattern involving downloading and sourcing a remote script bypasses local review of the script content before execution.
Evidence: SKILL.md Step 4 suggests: curl -sSLO https://raw.githubusercontent.com/open-edge-platform/edge-ai-libraries/main/microservices/model-download/scripts/get_model.sh followed by source ./get_model.sh.
[REMOTE_CODE_EXECUTION]: Automated scans detected patterns where network data is piped directly into a Python interpreter. While the analyzed instructions use the safe json.tool module for formatting or fixed local scripts for parsing, this pattern is a common attack vector and is flagged for its inherent risk.
Evidence: Multiple instances of curl ... | python3 -m json.tool and echo ... | python3 -c ... are present in SKILL.md and example files (e.g., examples/huggingface.md).
[EXTERNAL_DOWNLOADS]: The skill's primary function involves downloading large AI models from external sources such as HuggingFace, Ollama, and Ultralytics. This behavior is the intended primary purpose of the skill.

model-download-user