charles-review-checklist

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). High-risk signs of intentional obfuscation and a hidden runtime use of a private Objective‑C API (LSApplicationWorkspace via objc runtime + obfuscated string constants), plus hardcoded email-gated experiment logic that behaves like a per-account feature backdoor — these patterns suggest deliberate attempt to hide privileged behavior and enable special access for specific users.

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

• Potentially malicious external URL detected (high risk: 0.80). server/routes/openai.ts fetches the assistant instructions at runtime from the OpenAI Assistants API (asst_L4oto04GRe1IlhIA3q5rRNh6 — e.g. https://api.openai.com/v1/assistants/asst_L4oto04GRe1IlhIA3q5rRNh6), and that remote content is used to control/update prompts/instructions during execution.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The checklist explicitly references Stripe and subscription/billing behavior (e.g., "open the Stripe billing portal URL", "price IDs", plan-switch/billing logic, paywall/purchase lifecycle on PaywallManager/SuperwallManager). These are specific payment-gateway / subscription management concerns (Stripe) rather than generic developer tooling, so it meets the "Payment Gateways" criterion for Direct Financial Execution authority.