ZaStoGram_desktop/.claude/commands/rebase.md

description	argument-hint	allowed-tools
Drive an intent-aware rebase, resolving conflicts by reading both sides' history	onto <ref> | <branch> onto <ref> | tail of <branch> on top of <ref> | continue | abort	Bash, Read, Edit, Write, Grep, Glob, AskUserQuestion, TodoWrite

Intent-Aware Rebase

Read .agents/skills/rebase/SKILL.md completely and follow it with the substitutions below. Do not read .claude/ai-workflow-adapter.md: this is a branch-integration operation, not AI queue work, so none of that pipeline — task records, phase leaves, publication — applies.

Claude Code substitutions:

• Where the shared skill says to ask, use AskUserQuestion with the two intents in the question and 2–4 concrete resolution directions as options. Let it block: never continue the rebase past an unanswered question.
• Track the replay list with TodoWrite when it is more than a handful of commits, one item per commit, so a long rebase stays legible.
• The Bash tool has no interactive terminal, so the shared skill's non-interactive git rules are not optional here — git -c core.editor=true for every rebase --continue, and GIT_SEQUENCE_EDITOR for any todo-list edit. A command that opens an editor hangs until it times out.
• No delegation. Agent is deliberately absent from this command's tools: a resolution needs both sides' history in one context, and a subagent that returns a summary of what it read cannot supply that. Keep the reading targeted instead, per the shared skill's effort tiers.
• Keep stage dumps and the running report in the session scratchpad, never in the repository.

Request:

$ARGUMENTS