The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The install.sh script and README.md instructions include commands to fetch the uv installer from astral.sh and pipe it to the shell. This is a standard installation method for a reputable Python package manager from Astral Software and is treated as a safe external reference.
[COMMAND_EXECUTION]: The documentation and install.sh script involve commands that create directories and manage files in system-protected paths such as /root/clawd for OpenClaw integration. These operations are part of the intended administrative setup of the skill.
[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection as it retrieves untrusted data from Yahoo Finance, specifically via news articles and search results.
Ingestion points: Untrusted data enters the agent context through tool_get_news and tool_search_stocks in src/yfinance_mcp/tools/search.py.
Boundary markers: Absent. The skill returns strings within JSON objects but does not employ specific delimiters or instructions to the model to ignore embedded commands.
Capability inventory: The MCP server is limited to data retrieval and does not implement dangerous capabilities like arbitrary command execution, file system writes (outside of installation), or dynamic code evaluation.
Sanitization: The skill rounds numeric values and filters for nulls but does not sanitize textual content from news summaries or search results for potential injection patterns.

yfinance