ccopy

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly reads local Cursor transcripts and copies the last user+assistant turn verbatim to the clipboard/print output, which requires the LLM to include whatever text (including any secrets or API keys) appears in that turn in its output.

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

• Third-party content exposure detected (medium risk: 0.30). Source and runtime path: the skill reads outsider-authored free text from the local Cursor agent transcript JSONL files (or SQLite bubble text) under ~/.cursor/projects/*/agent-transcripts/*.jsonl via scripts/ccopy/transcript.py::parse_agent_transcript_last_turn / extract_message_text, which are arbitrary user/third-party submitted chat content without requiring prior selection of a specific item.