reflect-skills

Pass

Audited by Gen Agent Trust Hub on Jul 27, 2026

Risk Level: SAFEDATA_EXFILTRATION
Full Analysis
  • [DATA_EXFILTRATION]: The skill is designed to read and aggregate data from local session logs which contain sensitive information, including tool outputs and project-specific content. Evidence: File scripts/scan_sessions.py searches ~/.claude/projects/*/*.jsonl and ~/.codex/sessions/**/*.jsonl. However, the script processes this data locally, attempts to normalize error signatures to remove sensitive paths/hashes, and the skill instructions explicitly mandate discarding specifics and never copying raw transcript text or secrets.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 27, 2026, 03:06 AM
Security Audit — agent-trust-hub — reflect-skills