The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute an installation script for the Higgsfield CLI from the official GitHub repository of the vendor. This is documented as a standard setup procedure for the vendor's tools.
Evidence: curl -fsSL https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh | sh in SKILL.md.
[COMMAND_EXECUTION]: The skill uses the higgsfield CLI tool via Bash to perform authentication checks and trigger image generation jobs.
[INDIRECT_PROMPT_INJECTION]: The skill accepts user-provided prompts and images which are then passed to a backend system for processing. This presents a potential surface for indirect injection via untrusted user input.
Ingestion points: User-provided descriptions and image files collected during the pre-generation interview (Types A-F in SKILL.md).
Boundary markers: None explicitly defined in the CLI command arguments; instructions specify that the backend handles final prompt assembly.
Capability inventory: The skill executes shell commands (higgsfield product-photoshoot create) and accesses local files for image uploads.
Sanitization: The instructions state that the backend assembles the final prompt, which reduces the risk of direct injection by preventing the agent from generating the final prompt string.

higgsfield-product-photoshoot