text-analyst

Pass

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its core data processing workflow.
  • Ingestion points: The skill reads and analyzes raw text documents provided by the user in the 'data/raw/' directory (noted in phases/phase1-corpus.md and phases/phase3-analysis.md).
  • Boundary markers: No explicit delimiters or system instructions are defined to ensure the agent ignores instructions embedded within the corpus data.
  • Capability inventory: The agent has the capability to generate and execute R/Python scripts, perform shell operations via git (SKILL.md), and write files to the local system.
  • Sanitization: The skill does not implement sanitization or filtering of input text to detect or neutralize malicious instructions.
  • [COMMAND_EXECUTION]: The skill uses shell commands for project management and environment setup.
  • SKILL.md and phases/phase5-output.md describe using 'git add' and 'git commit' to manage the analysis workflow.
  • Technique guides instruct the installation of packages using 'pip install' and 'install.packages()', as well as downloading model data via 'python -m spacy download'.
  • [EXTERNAL_DOWNLOADS]: The skill references and downloads external datasets and software dependencies.
  • python-techniques/02_dictionary_sentiment.md contains a function to load the NRC Emotion Lexicon from saifmohammad.com.
  • R guides utilize the 'textdata' package to download standard sentiment lexicons like AFINN and NRC.
  • Guidance includes downloading data for NLTK (nltk.download) and spaCy models (en_core_web_sm).
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 1, 2026, 06:46 AM