hol-guard-protection

Pass

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill provides legitimate utility for managing and verifying security protections using the HOL Guard tool. The extensive documentation, including Architectural Decision Records (ADRs) and privacy contracts, consistently describes a security product focused on containment, redaction, and user approval of risky actions.
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to run commands whose output might contain data from external sources (e.g., package manager diagnostics or supply chain audit results). However, the primary purpose of the HOL Guard tool is to analyze and mitigate risks in such data.
  • Ingestion points: Agent context receives output from hol-guard commands such as hol-guard supply-chain scan and hol-guard command explain.
  • Boundary markers: The skill instructs the agent to review blocking reasons and suggests fix versions, emphasizing user oversight in the decision-making process.
  • Capability inventory: The skill uses hol-guard for subprocess evaluation, file system repair (shims), and application connection (e.g., Cursor editor integration).
  • Sanitization: The HOL Guard tool is explicitly documented to perform scanning and redaction of sensitive content (e.g., in command-activity-privacy.md and enterprise-evaluation-data-boundaries.md), serving as a defense against exfiltration and injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 14, 2026, 03:48 AM