The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to install a CLI tool using the command curl -fsSL https://cli.inference.sh | sh, which pipes a remote script from the vendor's domain directly to a shell for execution.
[EXTERNAL_DOWNLOADS]: During setup, the skill fetches binary distributions and verification checksums from the vendor's domain at dist.inference.sh.
[COMMAND_EXECUTION]: The skill relies on executing the infsh utility via the system bash shell to process image generation and search tasks.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by processing arbitrary user-provided text prompts and external image URLs.
Ingestion points: The prompt string and images URL array passed via the --input flag in SKILL.md.
Boundary markers: Absent; user-controlled data is interpolated directly into command-line arguments without delimiters or guardrail instructions.
Capability inventory: Performs shell command execution (infsh) and network-based image processing as defined in SKILL.md.
Sanitization: No input validation or URL sanitization is implemented for the data processed by the models.

nano-banana