audit-committee-support

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: No patterns of instruction override, safety filter bypass, or role-play injection were detected. The instructions are focused on providing structured support for audit committee tasks.
  • [DATA_EXFILTRATION]: The skill does not perform any network operations. While it attempts to read a local configuration file (../../CLAUDE.md), this is a common practice for accessing shared context in specific agent environments and does not target sensitive system files or credentials.
  • [COMMAND_EXECUTION]: There are no shell commands, subprocess calls, or scripts that would lead to arbitrary command execution.
  • [REMOTE_CODE_EXECUTION]: The skill does not download external scripts or packages. It is entirely composed of Markdown-based instructions and templates.
  • [INDIRECT_PROMPT_INJECTION]: While the skill is designed to process user-provided audit data, it lacks dangerous capabilities (like file writing or network access) that would make such injection exploitable. The primary output is formatted text.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 02:25 AM
Security Audit — agent-trust-hub — audit-committee-support