The Agent Skills Directory

[DYNAMIC_EXECUTION]: The calculator tool in the Basic Agent Graph section uses the eval() function to evaluate string expressions. This is a highly insecure practice as it allows for the execution of arbitrary Python code if the input string is not strictly controlled or sanitized.
[INDIRECT_PROMPT_INJECTION]: The skill architecture is susceptible to indirect prompt injection because the calculator tool processes inputs derived from the AgentState messages.
Ingestion points: User messages and potential outputs from the search tool are stored in the messages list within AgentState in SKILL.md.
Boundary markers: The example code lacks boundary markers or instructions to the LLM to ignore potentially malicious embedded content in data processed for tool calls.
Capability inventory: The skill possesses dynamic execution capability via the eval() call in the calculator tool.
Sanitization: There is no evidence of input validation, filtering, or sanitization before passing the string to the eval() function.

langgraph