The Agent Skills Directory

Remote Code Execution (CRITICAL): The skill contains a direct command curl -fsSL https://cli.inference.sh | sh in its Quick Start and usage examples. This pattern downloads a script from an external domain and executes it with shell privileges without any verification of the content. Since inference.sh is not a Trusted External Source, this represents a severe risk of arbitrary code execution.
Unverifiable Dependencies (MEDIUM): The skill suggests installing additional tools via npx skills add inference-sh/skills@.... This pattern encourages the user/agent to pull and execute code from an untrusted namespace, increasing the attack surface through dependency confusion or malicious package updates.
Indirect Prompt Injection (LOW): The skill uses infsh/agent-browser and tavily/search-assistant to process external web content.
Ingestion points: External URLs fetched via agent-browser and search results from tavily.
Boundary markers: Absent; there are no instructions to the agent to treat this data as untrusted or to ignore embedded instructions.
Capability inventory: The skill has Bash(infsh *) permissions, allowing it to perform network operations and local command execution.
Sanitization: Absent; the skill does not specify any filtering or validation for the content retrieved from the web before using it in thread generation.

twitter-thread-creation