The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill installs the higgsfield CLI by piping a remote script from GitHub directly to the shell (curl -fsSL https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh | sh). This execution method lacks integrity verification and allows for the execution of arbitrary code from an external source.
[EXTERNAL_DOWNLOADS]: The skill downloads software and configuration from a third-party GitHub repository. It also retrieves data from external URLs provided by the user to create marketing content, introducing unverified data into the agent's operating context.
[COMMAND_EXECUTION]: The skill relies on the Bash tool to run the Higgsfield CLI. It constructs shell commands using user-supplied inputs like prompts and URLs, which could be vulnerable to command injection if the agent or the CLI does not properly sanitize these inputs.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection (Category 8) by ingesting untrusted data from URLs and video analysis.
Ingestion points: Data enters through products fetch --url in SKILL.md and references/marketing-products.md, and via video analysis reports from the brain_activity model.
Boundary markers: No delimiters or instructions are used to separate external user data from system instructions.
Capability inventory: The skill has access to the Bash tool for file system access and network operations.
Sanitization: There is no evidence of filtering or validation for the fetched external content before it is processed by the agent.

higgsfield-generate