gpt2api-openai-gateway
Fail
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires cloning a repository from an unverified GitHub user (
https://github.com/432539/gpt2api.git). This source does not belong to a known trusted organization or the skill's stated author, increasing the risk of supply chain attack. - [REMOTE_CODE_EXECUTION]: The deployment instructions direct the user to run
docker compose up -d --buildon the cloned repository. This executes third-party code and builds containers from an untrusted source, which could contain malicious payloads or backdoors. - [CREDENTIALS_UNSAFE]: The core functionality of the skill involves 'reverse-engineering chatgpt.com' and 'multi-account pooling.' This requires users to provide sensitive ChatGPT
access_tokenandrefresh_tokenvalues. While the documentation mentions encryption, the software itself (from an unverified source) has full access to these credentials before encryption and at runtime. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts for image generation (
/v1/images/generations) and interpolates them into downstream requests to upstream AI services. - Ingestion points: User prompts in
v1/images/generationsandv1/images/edits(SKILL.md). - Boundary markers: None identified in the prompt interpolation logic described.
- Capability inventory: File writes (logs/backups), network operations (ChatGPT upstream, EPay integration, proxy pool), and Docker container execution.
- Sanitization: None described for external prompt content.
Recommendations
- AI detected serious security threats
Audit Metadata