The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from external DingTalk documents, which creates a surface for indirect prompt injection where malicious instructions embedded in documents could influence the agent's behavior.
Ingestion points: scripts/client.py (via get_document_data, download_file, and export_to_pdf).
Boundary markers: No delimiters or warnings are used when interpolating external content into the prompt context.
Capability inventory: The skill uses the httpx library for network operations and demonstrates file-writing capabilities in usage examples.
Sanitization: The scripts/transformer.py module performs basic noise filtering but lacks specific sanitization for prompt injection attacks.
[COMMAND_EXECUTION]: The documentation provides shell commands for dependency installation (pip, npx). While these are standard practices, users should ensure the installation source is trusted. The example scripts also show the agent writing downloaded binary content to the local filesystem.

dingtalk-docs-reader