The Agent Skills Directory

[DYNAMIC_EXECUTION]: The CalculatorTool implementation example in SKILL.md demonstrates the use of Python's eval() function to process string input provided by the agent. This pattern allows for arbitrary code execution within the environment if the agent's input is influenced by malicious instructions.
[INDIRECT_PROMPT_INJECTION]: The skill describes building agents that ingest data from external, untrusted sources, creating a potential vulnerability surface.
Ingestion points: Use of SerperDevTool for web search and ScrapeWebsiteTool for web scraping as described in SKILL.md.
Boundary markers: The provided examples do not use delimiters or explicit instructions to ignore embedded commands in the retrieved content.
Capability inventory: The agents are equipped with tools for network access, web scraping, file reading, and custom Python execution.
Sanitization: There is no evidence of filtering or sanitizing external content before it is interpolated into the agent's task context.
[EXTERNAL_DOWNLOADS]: The documentation includes instructions to install crewai and crewai[tools] using the standard Python package manager. These are well-known libraries for multi-agent orchestration.

crewai-agents