The Agent Skills Directory

Remote Code Execution (HIGH): The skill includes the command curl -fsSL https://cli.inference.sh | sh. This pattern downloads a script from an untrusted domain and executes it immediately with shell privileges, providing a direct path for arbitrary code execution on the host system.
Unverifiable Dependencies (MEDIUM): The skill uses npx skills add to fetch and install additional remote content from the inference-sh/skills repository. These dependencies are not from a trusted organization and represent an unverified supply chain risk.
Indirect Prompt Injection Surface (LOW): The skill uses infsh app run tavily/search-assistant to ingest external web data.
Ingestion points: Web search results via Tavily.
Boundary markers: Absent; the content is processed without clear delimiters to isolate untrusted data.
Capability inventory: The skill can execute bash commands and post to social media (X/Twitter), creating a path for malicious data to trigger actions.
Sanitization: No sanitization or validation logic is defined for the ingested search data.
Command Execution (LOW): The skill requests broad Bash(infsh *) permissions, allowing the agent to execute any sub-command of the custom infsh CLI tool.

linkedin-content