The Agent Skills Directory

[COMMAND_EXECUTION]: Documentation examples in references/tool-builder.md use the eval() function to process tool arguments (e.g., eval(call.args['expression'])). This pattern is insecure as it allows for arbitrary code execution if the input is not strictly validated or sanitized.
[COMMAND_EXECUTION]: The SKILL.md frontmatter specifies allowed-tools that include Bash(python *) and Bash(pip install inferencesh), granting the agent the ability to execute arbitrary Python scripts and install external packages in the environment.
[REMOTE_CODE_EXECUTION]: The skill documentation describes enabling an internal code_execution tool (references/agent-patterns.md and references/tool-builder.md) which permits the agent to write and execute code at runtime to solve problems.
[DATA_EXFILTRATION]: The SDK includes capabilities for uploading local files to remote storage via client.upload_file and attaching files to agent messages (references/files.md). This creates a vector for potential exfiltration of sensitive local data.
[DATA_EXFILTRATION]: The webhook_tool functionality documented in references/tool-builder.md allows the agent to send data to external HTTP endpoints, with examples specifically showing integration with services like Slack and GitHub.
[EXTERNAL_DOWNLOADS]: The skill instructions and code snippets frequently reference installing the inferencesh package from a public registry via pip install.

python-sdk