The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the user to run curl -fsSL https://cli.inference.sh | sh. This is a highly dangerous pattern that downloads a script from an external, non-trusted source and executes it immediately with the user's current shell privileges.
[EXTERNAL_DOWNLOADS]: The skill downloads binaries from dist.inference.sh during the installation process, and it also fetches various AI models and tool configurations from the inference.sh platform at runtime.
[COMMAND_EXECUTION]: The skill makes extensive use of a custom CLI tool (infsh) to execute commands for image generation, web searching, and application management. These commands are configured to run within a Bash environment as specified in the allowed-tools metadata.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its use of web-research tools.
Ingestion points: Untrusted data enters the agent context via the output of infsh app run tavily/search-assistant and infsh app run exa/answer as seen in SKILL.md.
Boundary markers: None are present to distinguish between internal instructions and external data retrieved from search results.
Capability inventory: The skill has shell execution capabilities restricted to the infsh command set.
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from external URLs before it is processed by the agent.

landing-page-design