The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs users and agents to install a CLI by piping a remote script directly to the shell: curl -fsSL https://cli.inference.sh | sh. This is a high-risk execution pattern as it runs unverified code from an external server on the host machine without prior review.
[COMMAND_EXECUTION]: The skill defines and uses the Bash(infsh *) tool to execute commands. These commands interact with the system to run AI models, manage tasks, and perform network-based operations.
[EXTERNAL_DOWNLOADS]: Both the automated installer and manual installation instructions download binary executables and checksum files from dist.inference.sh and cli.inference.sh. These sources are not on the pre-approved trusted organizations list.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests untrusted user input via the --input flag in commands like infsh app run, where prompt text is passed as JSON (e.g., in SKILL.md and references/running-apps.md).
Boundary markers: None identified. The instructions do not provide delimiters or instructions for the model to ignore potential injection attempts within the input data.
Capability inventory: The skill uses Bash to execute infsh, which communicates with external APIs and can perform actions like posting to Twitter/X.
Sanitization: There is no evidence of input sanitization or validation before the data is passed to the CLI tool.

agent-tools