The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions describe installing the belt CLI tool by downloading a script from https://cli.inference.sh and piping it to the shell (sh). This method executes remote code during the setup process. The skill also documents a manual installation method that supports SHA-256 checksum verification and Sigstore signature verification.
[INDIRECT_PROMPT_INJECTION]: The skill details how the belt CLI can automatically upload local files when a filesystem path is provided in place of a URL within the command inputs. This functionality creates an attack surface where an agent might be tricked by external data into exfiltrating local files.
Ingestion points: Inputs passed to the belt app run command via the --input flag, often in JSON format.
Boundary markers: The instructions do not define specific delimiters or warnings for the agent to ignore instructions embedded within these inputs.
Capability inventory: The skill utilizes the Bash tool to run the CLI, which has the capability to read files, perform network uploads, and execute subprocesses.
Sanitization: There are no documented sanitization steps for the input data before it is processed by the CLI tool.
[COMMAND_EXECUTION]: The skill requires the Bash tool to perform its primary tasks, allowing the agent to execute shell commands to manage AI apps and system configurations.

infsh-cli