The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documents an installation procedure for the official CLI using curl -fsSL https://cli.inference.sh | sh. This involves downloading and executing a script directly from the vendor's domain to configure the local environment.
[COMMAND_EXECUTION]: The skill makes extensive use of the belt CLI tool to perform operations such as executing AI applications, logging into the platform, and managing remote tasks.
[EXTERNAL_DOWNLOADS]: During installation, the skill fetches binaries and verification checksums from the vendor's distribution infrastructure (dist.inference.sh).
[INDIRECT_PROMPT_INJECTION]: The skill's primary functionality involves processing user-supplied JSON prompts and local files which are then transmitted to external AI models, creating a surface for indirect injection attacks.
Ingestion points: JSON input strings and local file paths passed to the belt app run command via the --input flag, as seen in SKILL.md and references/running-apps.md.
Boundary markers: None present. The instructions do not specify any delimiters or safety prompts to prevent the agent from obeying instructions embedded within the processed data.
Capability inventory: The skill is capable of reading local files, performing network uploads, and executing remote AI tasks through the inference.sh platform.
Sanitization: Not mentioned. Data is passed from the user or file system to the CLI tool without explicit sanitization or filtering steps documented.

agent-tools