The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent or user to install the CLI using the command curl -fsSL https://cli.inference.sh | sh. This pattern executes a script directly from a remote server controlled by the author.
[EXTERNAL_DOWNLOADS]: The installation process and manual setup instructions reference downloads for binaries, manifests, and checksums from dist.inference.sh and cli.inference.sh, which are vendor-owned domains.
[COMMAND_EXECUTION]: The skill defines allowed-tools: Bash(infsh *), which grants the agent the capability to execute any command provided by the infsh utility, including running models, managing tasks, and automating interactions with third-party services like Twitter/X.
[PROMPT_INJECTION]: The skill is exposed to indirect prompt injection risks because it processes results from untrusted AI applications and web searches which could contain malicious instructions.
Ingestion points: Data ingested through the --input flag in infsh app run and output returned from various AI 'apps' hosted on the platform.
Boundary markers: The CLI utilizes structured JSON for data exchange, which provides some structural separation but does not prevent the underlying AI models from obeying instructions embedded in the data.
Capability inventory: Full access to the infsh CLI tool via Bash allows the agent to perform a wide range of cloud-based operations.
Sanitization: The skill documentation does not define specific sanitization or filtering logic for data processed by the CLI before it is interpolated into agent prompts.

agent-tools