sn-ppt-entry
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on executing several localized Python scripts (
parse_user_docs.py,caption_images.py,launch_workbench.py) to process user attachments and manage the generation lifecycle. - [DYNAMIC_EXECUTION]: The orchestration logic involves generating and running dynamic Python snippets using
python3 -c. These snippets handle document data and facilitate communication with the SenseNova LLM for summarization purposes. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text from user-uploaded files, creating a vulnerability surface.
- Ingestion points:
scripts/parse_user_docs.pyextracts text from PDF, DOCX, Markdown, and TXT files. - Boundary markers: The
prompts/document_digest.mdfile defines a strict JSON output format to guide the LLM, although it lacks specific delimiters for the ingested text. - Capability inventory: The skill possesses the ability to execute shell commands, write persistent JSON files to the local filesystem, and make network requests to SenseNova APIs.
- Sanitization: Input data is truncated to a maximum character length in
scripts/parse_user_docs.pybefore being sent to the LLM. - [EXTERNAL_DOWNLOADS]: The skill includes checks for Node.js and may suggest that the user install it or other dependencies to enable the workbench features.
Audit Metadata