From next-step
Documents the next-step hook. Installs a UserPromptSubmit hook that injects a `**Next:**` directive on every user prompt so Claude ends each response with a single copy-pasteable slash command. No slash command of its own — it's pure hook behavior. Disable by uninstalling the plugin.
How this skill is triggered — by the user, by Claude, or both
Slash command
/next-step:next-stepThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
A tiny UserPromptSubmit hook that asks Claude to end every response with:
A tiny UserPromptSubmit hook that asks Claude to end every response with:
**Next:** `/some-command with args` — [brief description]
Reading a long response and then having to invent the next command yourself wastes attention. A **Next:** line at the bottom keeps continuity. This plugin is the lightweight version of that behavior — extracted from mk-flow's much bigger intent-inject.sh (259 lines of intent classification, rules engine, drift-check, etc.).
hooks/next-step.sh — the hook body. Prints 3 lines on stdout, which Claude Code injects into the user prompt.hooks/hooks.json — registers the hook on UserPromptSubmit.Edit hooks/next-step.sh. The text between the heredoc markers is exactly what Claude sees. Keep it terse — every byte is in your context window every prompt.
If you have Miltos's mk-flow plugin installed, BOTH this and mk-flow inject a **Next:** directive (mk-flow's includes much more — intent routing, rules, etc.). Pick one:
claude plugin uninstall mk-flow@mk-cc-resourcesclaude plugin uninstall next-step@protops-skillsnpx claudepluginhub panospro/protops-skills --plugin next-stepCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.