magic-linguistic-tokenize

Pass

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill's primary functionality is implemented through local Python scripts (fertility_audit.py, vocab_coverage.py) that perform calculations based on internal dictionaries and hardcoded baseline tables.
  • [SAFE]: No network-enabled commands (e.g., curl, wget) or external fetches are present in the runtime code. While documentation mentions external URLs for manual data refreshes (such as the SIL ISO 639-3 set), these are not automated by the skill.
  • [SAFE]: The skill processes text for tokenization analysis in sample_segmenter.py, but it lacks any sensitive capabilities (such as file-writing or subprocess execution) that could be exploited via malicious input text.
  • [SAFE]: All dependencies and references are consistent with the claimed linguistic purpose. The use of standard libraries like tiktoken for tokenizer analysis is appropriate and safe.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 3, 2026, 07:00 AM
Security Audit — agent-trust-hub — magic-linguistic-tokenize