caveman-compress

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly requires copying code blocks, inline code, URLs, commands, file paths, and environment variables "EXACTLY" which forces the model to reproduce any embedded API keys, tokens, or passwords verbatim if they appear in the input files.

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

• Third-party content exposure detected (high risk: 0.75). Outsider-authored free text can be read from a user-supplied file path at runtime (e.g., /caveman-compress <filepath> → compress_file() → filepath.read_text()), and that file’s contents are then embedded into the LLM prompt via build_compress_prompt(body) → call_claude(... messages=[{"role":"user","content": prompt}]).