deployment-record-archive

Warn

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • Indirect Prompt Injection (MEDIUM): The skill processes deployment records from external sources and appends them to a persistent JSONL archive, creating a risk of log poisoning.\n
  • Ingestion points: Untrusted data enters via record.json in scripts/archive_record.py.\n
  • Boundary markers: None present; data is directly serialized.\n
  • Capability inventory: File write (append) via open(sys.argv[2], 'a') in scripts/archive_record.py.\n
  • Sanitization: Limited to standard JSON serialization; no content-based filtering of metadata fields.\n- Command Execution (LOW): The Python script accepts file paths as command-line arguments without validation. While the skill's defined command uses relative paths, a manipulated agent could potentially be coerced into reading or writing sensitive files elsewhere on the system.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 16, 2026, 11:05 AM