doc-intel-llm-tier-pl
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill relies on the
contextgemlibrary (version 0.25.1 or higher), which is a third-party package that must be downloaded from the Python Package Index (PyPI). While necessary for the skill's functionality, it introduces an external dependency beyond the standard library environment. - [COMMAND_EXECUTION]: The documentation provides explicit shell commands for installing dependencies (
pip install contextgem>=0.25.1) and executing the extraction scripts. The main scriptscripts/ekstrakcja_llm.pyuses command-line arguments to specify file paths and model parameters. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external legal documents and interpolates their content into LLM prompts for concept extraction, creating a surface for indirect prompt injection.
- Ingestion points: Reads content from
.txtand.docxfiles specified via command-line arguments inscripts/ekstrakcja_llm.py. - Boundary markers: The script does not implement explicit boundary markers or warnings to the LLM within its own code, relying on the underlying
contextgemlibrary for prompt construction. - Capability inventory: The skill has the capability to read local document files and make network requests to both local (Ollama) and cloud-based LLM APIs (OpenRouter, OpenAI) as seen in
scripts/ekstrakcja_llm.py. - Sanitization: There is no evidence of input filtering or sanitization performed on the document text before it is passed to the LLM processor.
Audit Metadata