promptifylisted

# Prompt upgrade Turn the raw prompt below into the prompt an expert on this codebase would have written for this model: specific, grounded in real files, verifiable, and no longer than the task needs. The rewritten prompt is the deliverable; the task itself is not executed unless `--run` is passed. ## Input <raw_prompt> $ARGUMENTS </raw_prompt> If `<raw_prompt>` is empty, treat the user's previous message as the raw prompt. If there is none, ask for it and end the turn. Flags (strip them from the raw prompt before working): - `--run` — after presenting the upgraded prompt, execute it as written in this same turn. - `--model <name>` (also accepted as `--mode <name>`) — tune for that model instead of the one running this session (for example a subagent on Sonnet 5, or an API app on Opus 5). Without it, the target is the current session model, which your system prompt names. ## Repo snapshot Collected automatically before you start (session effort: `${CLAUDE_EFFORT}`): !`${CLAUDE_SKILL_DIR}/scripts/snapshot.sh` ## Workflow 1. **Classify.** Decide, briefly, what kind of prompt this is: feature, bug fix, refactor, question or explanation, code review, research or investigation, docs or writing, design or frontend, a long autonomous run, or a system prompt for an application. Decide its size: a one-sentence diff, a bounded change, multi-file work, or autonomous. And decide what it really asks for: implement, suggest, or explain. These three choices drive everything belo