The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install a CLI tool by cloning a repository from an unverified source (github.com/Ito-Markets/ito-cloud-runtime.git) and running npm ci. This process involves executing code from a repository that is not recognized as a trusted vendor.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external GPU inventory queries and RFQ status updates through its tool interfaces, which could be exploited to inject instructions.
Ingestion points: Untrusted data enters the agent context via the output of the ecc ito find and ecc ito status commands as described in SKILL.md.
Boundary markers: The skill does not provide delimiters or instructions for the agent to distinguish between its own system instructions and the data returned from external commands.
Capability inventory: The skill allows the execution of multiple subprocesses using the ito CLI, including find, accept, and evals commands.
Sanitization: There is no evidence of sanitization or filtering of the output received from the external API tools.
[DYNAMIC_EXECUTION]: The skill is configured to execute JavaScript files (ito.js, ito-mcp.js) at runtime based on paths provided through environment variables like ECC_ITO_CLI_EXECUTABLE. This allows for the dynamic loading and execution of code from potentially arbitrary locations on the file system.

ito-compute