USE WHEN the user asks to format references for an academic paper or research note. Converts messy DOIs, URLs, or pasted text into APA / IEEE / GB/T 7714 / BibTeX consistently. Verifies the cite key matches the work.
USE WHEN the user pastes code or asks for review of a file/diff. Reads the code, identifies real issues (bugs, security, perf, correctness, readability) in priority order, suggests concrete fixes. Skips style nitpicks unless asked.
USE WHEN a user shares or references a CSV file and wants it summarized, analyzed, or visualized — produces data overview, summary statistics, missing-data report, and relevant charts using built-in tools (Read + Bash running Python/pandas).
USE WHEN the user asks to clean up, reformat, or polish a markdown document. Normalizes headings, lists, tables, code fences, math delimiters, and Chinese punctuation. Returns the rewritten doc, not a description of changes.
USE WHEN turning raw meeting notes, transcripts, or discussion logs into structured minutes — extracting summary, decisions, action items (with owners and due dates), and next steps using ready-made templates for standups, project reviews, and client calls.
USE WHEN the user wants to draw an architecture diagram, sequence diagram, flowchart, ER diagram, or state machine. Picks the right mermaid type, writes correct syntax, validates before returning.
USE WHEN a user wants to create a PowerPoint (.pptx) presentation or slides — generates a deck programmatically with python-pptx, adding a title slide, content slides with headings and bullet points, and saving the result as a .pptx file.
USE WHEN the user pastes long text (article, paper, transcript, doc) and wants a summary. Picks the right summary shape (TL;DR / key points / structured / actionable) for the source type and length. Preserves numbers, names, dates verbatim.