Cursor Chat Prompt Analysis

Problem it addresses

Cursor chats often span multiple Conversation entries, and the same bug may be debugged across separate turns without a clear prompt review trail. This skill turns Cursor chat .md logs from a given date or directory into an xlsx report organized by distinct issues.

How it works

It first parses date conditions such as 'yesterday', 'last week', or '2026-04-20', then looks for {date}_{hash}.md files in the local cursor_logs folder. If no matching file exists, it invokes cursor-chat-export for that date and asks whether to step backward day by day when no record is found. It then checks Python and openpyxl compatibility for >=3.1.0, reads the Markdown logs, and identifies independent issues by topic shifts, transitions, shared files or modules, and prefixes such as fix:, feat:, or ana:.

Each report row represents one independent issue and includes: issue summary, conversation count, initial prompt, final resolution prompt, key solution factor, prompt strategy, optimization suggestions, and notes. The skill preserves the full original first user message and final key user message, which is useful for reviewing strategies like context referencing, step-by-step guidance, error feedback, iterative correction, explicit constraints, and task prefixes. After the table, it writes a summary section containing general prompt optimization suggestions and reusable general prompt rules.

Boundaries

This is best for engineers or developers reviewing their own Cursor prompting history. It assumes chats have been exported to Markdown and relies on cursor-chat-export plus a suitable Python environment. It analyzes existing conversation boundaries and prompt quality; it does not generate code inside Cursor or modify workspace files automatically. If a selected date has no records, the user must confirm whether to check earlier dates.