spreadsheet-analysis

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md workflow step 1, the agent opens a user-provided spreadsheet/CSV file path and runs the bundled profiler (scripts/profile_table.py) at runtime, which ingests that outsider-authored file contents (e.g., CSV records and some XLSX formula text) into LLM-adjacent JSON evidence.