genshijin-compress

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill sends raw file contents to the Anthropic/Claude API and requires exact preservation of code blocks/inline code (which may contain secrets), so the LLM would receive and could output secret values verbatim despite filename-based exclusions.

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

• Third-party content exposure detected (medium risk: 0.30). ユーザーが指定した任意のローカル filepath の内容を scripts/compress.py が original_text = filepath.read_text(...) で読み取り、その全文を call_claude(build_compress_prompt(original_text)) により Anthropic/Claudeへ送って圧縮（またはエラー時の修正）しているため、外部（ユーザー起点）の自由記述テキストがそのままLLM入力に混入し得ます。