The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends the installation of external components via npx skills add, specifically targeting belt-sh/cli and various sub-skills under the inference-sh/skills GitHub repository. It also links to external documentation and installation guides hosted on raw.githubusercontent.com/inference-sh/.
[COMMAND_EXECUTION]: The skill relies on the belt CLI utility for its primary functionality. It provides instructions for authentication (belt login), listing applications (belt app list), and executing specific image generation models (belt app run <model> --input ...). The allowed-tools frontmatter correctly scopes access to this specific command pattern.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests user-supplied text to pass to external image generation models.
Ingestion points: User input is interpolated into the --input JSON payload for the belt command (found in SKILL.md).
Boundary markers: The skill uses structured JSON formatting to encapsulate user prompts, which provides a natural boundary for the external command.
Capability inventory: The skill is limited to shell execution of the belt tool via the Bash tool.
Sanitization: There is no explicit sanitization shown in the markdown examples, but the use of JSON structures minimizes the risk of command injection within the local shell context.

ai-image-generation