The Agent Skills Directory

REMOTE_CODE_EXECUTION (CRITICAL): The skill utilizes the highly dangerous pattern curl -fsSL https://cli.inference.sh | sh to install its dependencies. This allows an external server to execute arbitrary code on the local system without verification. Additionally, the skill uses npx skills add which downloads and executes packages from the npm registry, further increasing remote code execution risk.\n- EXTERNAL_DOWNLOADS (HIGH): The skill downloads and executes code from inference.sh and the npm registry, neither of which are in the trusted sources list. This poses a significant supply chain risk.\n- COMMAND_EXECUTION (HIGH): The skill performs shell command execution using the Bash tool. It interpolates user-provided prompts into JSON strings passed as arguments to the infsh binary, creating a high risk of command injection if the user provides malicious input that breaks the JSON or shell syntax.\n- PROMPT_INJECTION (LOW): The skill is vulnerable to indirect prompt injection due to a lack of sanitization for user-provided inputs used in tool calls.\n
Ingestion points: The prompt field within the --input argument of infsh commands in SKILL.md.\n
Boundary markers: None present to delimit user input or warn the agent about embedded instructions.\n
Capability inventory: File system access and network operations via the infsh binary and Bash shell.\n
Sanitization: No sanitization or validation of the input prompt is performed before interpolation into the shell command.

ai-product-photography