The Agent Skills Directory

REMOTE_CODE_EXECUTION (CRITICAL): The skill documentation includes a command to download and execute a script directly from a URL: curl -fsSL https://cli.inference.sh | sh. This is a highly dangerous pattern as it executes unverified code with the privileges of the current user.
Evidence: Source URL: https://cli.inference.sh.
Trust Status: Untrusted. The domain inference.sh is not included in the allowed list of trusted organizations or repositories.
EXTERNAL_DOWNLOADS (MEDIUM): The skill utilizes npx skills add to fetch and install external skills from the inference-sh organization. These dependencies are not version-locked and reside on an untrusted external registry.
INDIRECT_PROMPT_INJECTION (LOW): The skill uses the infsh tool to perform web searches (e.g., tavily/search-assistant, exa/search), which ingests untrusted third-party data into the agent's context.
Ingestion points: Output from infsh app run commands (search results and market data).
Boundary markers: Absent. The skill does not define delimiters or instructions to ignore embedded commands in the search output.
Capability inventory: The allowed-tools includes Bash(infsh *), which provides broad access to execute various CLI applications and system commands.
Sanitization: Absent. There is no evidence of filtering or escaping logic for the data returned from search providers.

press-release-writing