The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill is designed to process external machine learning code and configurations while requesting access to powerful tools including Bash, Write, and Edit. There is no evidence of input sanitization or boundary markers to prevent malicious instructions embedded in ML code from manipulating the agent's actions.
Ingestion points: Reads user-provided ML scripts and deployment configurations (implied by Purpose/Capabilities).
Boundary markers: Absent. The skill does not define delimiters for untrusted content.
Capability inventory: Bash, Write, Edit, Read, Grep tools are enabled.
Sanitization: None described. The agent is encouraged to generate and validate code based on these inputs.
[Command Execution] (MEDIUM): The skill explicitly requests the Bash tool. While no malicious shell commands are hardcoded, the capability to execute arbitrary shell commands on a system is highly dangerous when combined with the processing of untrusted external data (ML models/scripts).
[No Code] (INFO): This skill contains no executable scripts or binaries within its own definition; it relies entirely on instructions and the agent's pre-existing toolset.

torchscript-exporter