meccha-chameleon-game-trainer-detection
Fail
Audited by Gen Agent Trust Hub on Jul 1, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Provides a link to an external archive hosted on Netlify (https://skydock.netlify.app/trainer-archive.zip) which the documentation describes as a potential malware distribution vector.
- [COMMAND_EXECUTION]: Includes Python code snippets utilizing the ctypes library to interact with sensitive Windows APIs such as OpenProcess, WriteProcessMemory, and CreateRemoteThread for memory manipulation and DLL injection.
- [REMOTE_CODE_EXECUTION]: Instructions direct the agent to download and analyze executable files, which presents a risk of executing untrusted code if performed without robust isolation.
- [PROMPT_INJECTION]: The skill processes content from untrusted external repositories, creating an attack surface for indirect prompt injection.
- Ingestion points: Analyzes data from external game trainer repositories.
- Boundary markers: No instructions or delimiters are provided to isolate external content from the agent's core logic.
- Capability inventory: Provides code templates for low-level system access, network communication, and process management.
- Sanitization: No validation or sanitization of ingested content is documented.
Recommendations
- AI detected serious security threats
Audit Metadata