bounded-agency-review

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and review potentially untrusted AI agent instructions, prompts, and configuration files, creating a surface for indirect prompt injection.
  • Ingestion points: The skill processes user-supplied text files including SKILL.md, prompts, and workflows (SKILL.md, Step 1).
  • Boundary markers: The skill requires a mandatory 'Material Coverage' declaration (SKILL.md, Step 2) to explicitly define the boundaries of what was analyzed and what remains unverified.
  • Capability inventory: The skill is strictly limited to generating markdown reports and does not possess capabilities for network operations, code execution, or privileged file access.
  • Sanitization: No explicit sanitization or escaping mechanisms for untrusted inputs are defined.
  • [SAFE]: The skill logic encourages strict boundaries and audit trails. No evidence of malicious obfuscation, remote code execution, persistence, or data exfiltration was found in the provided instructions or guidelines.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 05:41 AM
Security Audit — agent-trust-hub — bounded-agency-review