The Agent Skills Directory

Command Execution (SAFE): The skill provides instructions for standard development commands like reflex run and reflex init. These are appropriate for the framework's operation and described use cases.
External Downloads (SAFE): The skill recommends installing reflex, which is a well-known and legitimate Python web framework. Per [TRUST-SCOPE-RULE], these downloads are considered safe as they belong to a trustworthy ecosystem.
Indirect Prompt Injection (LOW): \n
Ingestion points: The script examples/file_upload.py accepts user-uploaded files through the rx.UploadFile component. \n
Boundary markers: No explicit markers or instructions are provided to the agent to treat filenames or content as untrusted. \n
Capability inventory: The skill uses open() to write files to the local disk in the ./uploads/ directory (file: examples/file_upload.py). \n
Sanitization: While the code validates file extensions, it does not sanitize the file.filename attribute for path traversal characters (e.g., ../), creating a theoretical vulnerability surface where a malicious filename could target sensitive paths outside the intended directory.
Data Exposure & Exfiltration (SAFE): Investigated the automated scan alert for 'state.in'; no literal URL or malicious usage of this domain was found in the provided files. The alert likely triggered on substrings within legitimate code, such as State.increment or State.items.

reflex-dev