The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script scripts/extrair_media.py using the Python interpreter. The script is written using only standard library modules to avoid third-party dependency risks.\n- [EXTERNAL_DOWNLOADS]: The script downloads content from Google's export endpoint, restricted to docs.google.com and *.googleusercontent.com. It validates response headers and file signatures to ensure only legitimate presentation files are processed.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from URLs and uploaded files. It uses explicit boundary instructions to treat content as data and restricts extraction to binary media folders, preventing the processing of embedded text as instructions. Ingestion points include Google Slides URLs and PowerPoint files; capability inventory includes file system writes and network reads; sanitization includes path validation and signature checking.\n- [DATA_EXFILTRATION]: Network activity is monitored and restricted to verified Google domains with additional checks on server identities to prevent the tool from being used as a proxy for unauthorized data transmission.

extrair-imagens