text-analyst

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [Indirect Prompt Injection] (LOW): The skill is designed to ingest and analyze untrusted text data from external files in the 'data/raw/' directory. Malicious instructions embedded in the corpus could potentially influence agent behavior during tasks involving reading and summarizing document content.\n
  • Ingestion points: Local file system access to CSV and text files in 'data/raw/' as specified in 'phases/phase1-corpus.md' and 'phases/phase4-validation.md'.\n
  • Boundary markers: Absent. The skill does not implement specific delimiters or 'ignore' instructions when processing corpus text.\n
  • Capability inventory: The agent performs file write operations (memos/, output/) and executes complex data analysis code in R and Python.\n
  • Sanitization: No sanitization or safety-filtering logic is applied to the content of the analyzed documents.\n- [External Downloads] (SAFE): The technique guides recommend installing several standard industry-standard libraries (e.g., 'nltk', 'bertopic', 'tidytext') from reputable registries like CRAN and PyPI. These are expected dependencies for the skill's stated purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:39 PM