ai-director

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt includes examples and commands that require embedding API keys or verification codes verbatim (e.g., bind --key "x2c_sk_xxx", credentials/{USER_ID}.json with "x2cApiKey"), so an agent would need to read and output secret values directly, creating an exfiltration risk.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The codebase contains high-risk patterns: a hard-coded, non-obvious external API endpoint receiving user API keys and account actions, plus logic that writes unified account data and attempts to run a sync script (child_process.execSync), which together create a plausible credential-exfiltration/backdoor vector and remote-code-execution opportunity.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill's quality-evaluator calls the Gemini REST endpoint at https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent?key=${GEMINI_API_KEY} at runtime and uses the model's JSON response (including "prompt_improvements") to modify the generation prompt in auto-iterate, so remote content directly controls subsequent agent prompts and the workflow requires the Gemini API key.