read-line-messages

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs the transcriber to "transcribe every message verbatim" (a verbatim-constrained vision pass), which forces the LLM to output any secrets (passwords, API keys, tokens) present in the LINE screenshots verbatim, creating an exfiltration risk.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The code is explicitly designed to capture private LINE conversation screenshots and pass them to a "codex" transcription agent (via codex exec), which creates a high risk of intentional data exfiltration to an external service even though there is no obfuscated backdoor or credential-theft code present.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). Source scripts/run.ts captures LINE chat contents via screencapture -l<windowid> from the user’s currently open/outside-authored conversation and then the transcribe codex pass ingests that screenshot text at runtime.