The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill explicitly instructs users to install external components using the npx skills add command, targeting belt-sh/cli and multiple repositories under the inference-sh organization. These sources are not recognized as trusted providers, which poses a supply chain risk by executing unverified remote code during the skill's setup and operation.
[COMMAND_EXECUTION]: The instructions include multiple Bash code blocks that invoke an external CLI tool (belt). Although the allowed-tools configuration in the frontmatter attempts to limit Bash to only the belt command, the execution of this external binary involves processing data and communicating with a remote service (inference.sh), which is outside the local safety boundary of the agent.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill takes user-defined prompts for image and video generation as seen in the belt app run examples.
Boundary markers: No markers or delimiters are used to separate user-provided data from instructions within the shell command arguments.
Capability inventory: The skill possesses the capability to execute shell commands via the Bash tool.
Sanitization: There is no evidence of sanitization or input validation for the prompts before they are passed to the external CLI tool, allowing a surface for potentially malicious instructions to be processed by the underlying AI models.

app-store-screenshots