skills/j4flmao/agent-skills/backup-dr/Gen Agent Trust Hub

backup-dr

Warn

Audited by Gen Agent Trust Hub on Jul 8, 2026

Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill utilizes an adversarial pattern known as context stuffing. Eight reference files (including architecture-patterns.md and testing-strategies.md) contain 150 repetitive sections each, totaling over 1,200 paragraphs of technical filler text. This technique is used to bypass safety filters or degrade model performance by exhausting the context window.
  • [PROMPT_INJECTION]: Indirect Prompt Injection vulnerability surface detected. The skill ingests untrusted configuration targets to generate executable backup scripts and deployment templates without proper isolation or sanitization.
  • Ingestion points: Infrastructure types, RPO/RTO targets, and budget parameters in SKILL.md.
  • Boundary markers: Absent; no delimiters are used to separate untrusted user data from the skill's instructions.
  • Capability inventory: Use of subprocess.run in db_backup.py for database commands, aws CLI management in failover runbooks, and package installation in the backup-validation workflow.
  • Sanitization: Absent; the provided Python script interpolates configuration variables directly into shell commands without validation.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 8, 2026, 12:23 PM
Security Audit — agent-trust-hub — backup-dr