취약점

-= Per source details. Do not edit below this line.=-

Source: amazon-inspector (1d53e4571f8ccfc385a265dfd47cbea9793946762a794aff432e98614ee10b21)

The package ships epl/.ai_config.json containing a hardcoded Groq API key with provider set to 'groq'. On any AI-related CLI invocation (epl ai, epl gen, epl explain <file>) or call to epl.ai.generate/chat, ai._load_config() reads this bundled config from the install directory and _use_cloud() returns True. The user's prompts — which for epl explain <file> include the contents of the user's source files — are then POSTed to https://api.groq.com/openai/v1/chat/completions authenticated with the bundled key (epl/ai.py line 262: 'Authorization': f'Bearer {CLOUD_API_KEY}'). Because the key belongs to the package author's Groq account, every conversation and every file the user asks the tool to explain becomes visible in the author's Groq dashboard. There is no README disclosure, no opt-in prompt, and no indication to the user that their data is leaving their machine via author-controlled credentials. This is the silent-relay shape: the package's advertised AI feature has a hardcoded destination + author credential, so normal use of the API leaks caller-supplied data — including potentially proprietary source code — to a third party the user did not knowingly authorize.