The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions frequently use the eval "$(python scripts/xxx.py ...)" pattern to execute Python scripts and then evaluate their output in a shell. This pattern is inherently risky as it executes output directly in the shell environment.
[DYNAMIC_EXECUTION]: The Python scripts (scripts/search.py, scripts/enum-params.py) are used to dynamically generate JavaScript code blocks. These generated blocks are intended to be executed within a browser environment via browser-act. The scripts/search.py script specifically interpolates user-supplied search queries into a JavaScript string template using .format(), which is a form of dynamic code generation.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface by fetching search results from an external API (Zhihu) and parsing them into a structured format for the agent's context.
Ingestion points: Data enters the context via the results array in the JSON output of the search API call in scripts/search.py.
Boundary markers: None observed in the Python output or the markdown instructions for handling the API response.
Capability inventory: The skill uses eval in shell and generates JS for browser execution, providing significant capabilities.
Sanitization: Basic escaping of single quotes is performed for the search query in scripts/search.py, but there is no evidence of sanitization for the data retrieved from the API before it is processed by the agent.

zhihu-search