ingest-cwe-taxonomies

Pass

Audited by Gen Agent Trust Hub on Jul 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADS
Full Analysis
  • External Resource Access: The skill is configured to fetch the official CWE database from MITRE's servers (cwe.mitre.org). This is a well-known and reputable source for security metadata.
  • Indirect Prompt Injection Surface: The skill ingests and processes external XML data to generate taxonomies. While the source is a standard security repository, the ingestion of external corpora is an inherent characteristic of data processing skills.
  • Ingestion points: MITRE's XML repository and user-provided local paths.
  • Boundary markers: The skill workflow refers to a secondary prompt file for specific instruction boundaries.
  • Capability inventory: The skill performs file system writes to taxonomies/, data/, and scripts/ directories, and updates a manifest file.
  • Sanitization: Sanitization logic is likely contained within the referenced ingestion script and prompt instructions.
  • Script Generation: The workflow generates a Python ingestion script (scripts/ingest-cwe.py) to handle the normalization and mapping phases, which is a common pattern for automating data transformations.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 10, 2026, 02:50 PM
Security Audit — agent-trust-hub — ingest-cwe-taxonomies