The Agent Skills Directory

[COMMAND_EXECUTION]: The implementation of the calculator tool in SKILL.md uses the Python eval() function to process input strings. This is a significant security risk as it allows for the execution of arbitrary Python code if the agent is provided with a malicious expression.
[COMMAND_EXECUTION]: The troubleshooting section of SKILL.md includes the run_notebook function, which uses nbconvert.preprocessors.ExecutePreprocessor to programmatically execute all code cells within a Jupyter notebook.
[EXTERNAL_DOWNLOADS]: The skill's setup instructions direct users to clone an external repository from GitHub (github.com/FareedKhan-dev/all-agentic-architectures.git).
[PROMPT_INJECTION]: The architectural patterns in SKILL.md create a surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context through parameters such as user_input, task, and problem across multiple patterns including ReflectionState, MultiAgentState, and ToTState.
Boundary markers: Prompt templates do not use delimiters or explicit instructions to ignore embedded instructions within processed data.
Capability inventory: The skill provides high-risk capabilities, including eval()-based code execution, notebook execution, and network-enabled search tools.
Sanitization: There is no evidence of input validation, escaping, or filtering before untrusted content is interpolated into agent prompts.

all-agentic-architectures