sentinel-ai-security
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [EXTERNAL_DOWNLOADS] (HIGH): The skill instructs users to install code from an untrusted GitHub repository.
- Evidence: SKILL.md directs users to clone 'github.com/DmitrL-dev/AISecurity.git'.
- Risk: This source is not verified or trusted, potentially leading to remote code execution.
- [COMMAND_EXECUTION] (HIGH): The skill includes capabilities for network attacks and kernel-level traffic interception.
- Evidence: SKILL.md mentions 'sentinel strike' with 39,000+ payloads. workflows/context.md mentions 'sentinel-driver' for WFP traffic interception.
- Risk: Use of kernel drivers and offensive security tools can compromise host security and facilitate malicious activity.
- [PROMPT_INJECTION] (LOW): Indirect Prompt Injection vulnerability surface.
- Ingestion points: Analysis of untrusted user strings via 'sentinel analyze' in SKILL.md.
- Boundary markers: None documented in instructions.
- Capability inventory: Automated network attacks (strike), Kernel driver (sentinel-driver), network requests.
- Sanitization: No sanitization logic provided in the skill instructions.
- [CREDENTIALS_UNSAFE] (LOW): Presence of default test credentials.
- Evidence: workflows/dev.md includes 'password: sentinel' for an administrative login test.
Recommendations
- AI detected serious security threats
Audit Metadata