privileged-access

Pass

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious code or suspicious behavior detected. The skill consists of markdown-based instructions for security assessment.
  • [PROMPT_INJECTION]: Benign instructional patterns detected. The skill contains the phrase "ignore previous instructions" within an "Injection Hardening" section. This is a defensive guideline instructing the agent to detect and report injection attempts in target data rather than a malicious payload.
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill processes untrusted external data, but includes robust safeguards.
  • Ingestion points: Files and directories specified in arguments, including PAM configurations, vault metadata, and session recordings.
  • Boundary markers: Includes a dedicated "Injection Hardening" section with explicit instructions to ignore directives found in input data and flag them as findings.
  • Capability inventory: Uses read-only tools (Read, Grep, Glob) for data analysis.
  • Sanitization: Relies on system-level instructions to ignore and report embedded commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 15, 2026, 01:07 AM
Security Audit — agent-trust-hub — privileged-access