striplog
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources which could potentially contain malicious instructions that an LLM might misinterpret as system commands.
- Ingestion points: The
scripts/create_striplog.pyscript reads data from external CSV files usingpd.read_csv()and parses raw text descriptions viaStriplog.from_description(). - Boundary markers: There are no explicit delimiters or boundary markers defined in the instructions to isolate the parsed data from the agent's internal instruction context.
- Capability inventory: The skill has file-writing capabilities through
matplotlib.savefig()for visualization andstriplog's export functions for CSV, JSON, and LAS files, which could be abused if an injection is successful. - Sanitization: The skill relies on the
striploglibrary's parsing logic but does not implement additional sanitization or filtering for the input text or file content.
Audit Metadata