The Agent Skills Directory

REMOTE_CODE_EXECUTION (CRITICAL): The skill contains a 'Quick Start' instruction to run curl -fsSL https://cli.inference.sh | sh. This is a highly dangerous pattern that downloads and executes an arbitrary script from an external server with the privileges of the current shell. This domain is not on the list of trusted external sources.\n- EXTERNAL_DOWNLOADS (MEDIUM): The skill promotes the addition of other skills from the inferencesh/skills repository using npx. These components are not from a verified or trusted organization and could lead to the execution of malicious logic within the agent's environment.\n- PROMPT_INJECTION (LOW): The pipeline definitions are vulnerable to indirect prompt injection as they ingest external data (prompts, URLs, script text) and interpolate them directly into tool calls.\n
Ingestion points: prompt, text, image_url, and video_url fields within the infsh app run command payloads in SKILL.md.\n
Boundary markers: None present; data is passed directly as string values within JSON inputs.\n
Capability inventory: The skill uses the Bash tool to execute the infsh CLI, which has capabilities to generate, modify, and merge media files.\n
Sanitization: No sanitization or validation of the input strings is performed before they are passed to the external CLI tool.\n- COMMAND_EXECUTION (LOW): The skill relies on executing system commands through the Bash tool. While restricted to the infsh binary in the allowed-tools field, the reliance on shell execution for core functionality increases the attack surface if input handling is flawed.

ai-content-pipeline