The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/upload_to_dataset.py executes the curl binary via the subprocess.run function to perform all network operations. While this uses an argument list rather than a shell string, it remains a method of system command execution.
Evidence: The run_curl_json function in scripts/upload_to_dataset.py is invoked throughout the workflow (discovery, upload, pipeline execution, and metadata management) to interact with the Dify API endpoints.
[INDIRECT_PROMPT_INJECTION]: The skill processes external file data, creating an attack surface where malicious instructions embedded in files could be processed by the agent or the target system.
Ingestion points: Local files are read via the --file CLI argument in scripts/upload_to_dataset.py and transmitted to the Dify API.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore or isolate content within the files being uploaded.
Capability inventory: The skill possesses file read capabilities and network access via subprocess calls to curl.
Sanitization: The script does not perform sanitization or validation of the file content before transmission to the remote API.

dify-knowledge-base-upload