Slash Command System Prompts as a Personal AI Interface

The Insight

A well-designed system prompt is not just a set of behavioral instructions — it can include a personal shorthand vocabulary that turns a general-purpose AI into a responsive, task-aware assistant tuned to your specific workflow. By embedding named slash commands directly into the system prompt, you create a persistent interface layer that collapses multi-step instructions into single tokens.

Lou shared his full production system prompt in the September 18 session chat. It combined a behavioral brief with a complete command library, effectively giving Claude a menu of pre-defined operations the user could invoke by name.

The Framework

Part 1 — Behavioral Brief: Sets the working relationship: conversational but rigorous, concise unless asked for detail, willing to challenge as well as affirm, oriented toward real outcomes.

“Speak to me in a conversational, thoughtful tone — like a smart, capable, persistent assistant who won’t stop until we’ve cracked the problem. Be concise but thorough: I value clarity and insight over verbosity… I like support and challenge: affirm what’s working in my thinking, and pressure-test it with alternatives or blind spots I may not see.”

Part 2 — Slash Command Library:

CommandFunction
/frame [topic]Structure or reframe a complex issue
/gutcheck [decision/options]Reflect choices to guide intuitive clarity
/challenge [idea]Surface blind spots and test assumptions
/simplify [concept]Make complex ideas clear and teachable
/outline [topic]Create a curiosity-driven, hook-rich outline
/punchup [text]Enhance engagement, voice, and flow
/optimize [goal] [draft]Refine content for a specific outcome
/aha [topic]Surface a novel or insight-triggering angle
/tonecheck [text]Calibrate for voice and tone
/hookify [intro]Add curiosity or narrative pull
/compare A vs BCrisp pros/cons contrast
/scan [draft]Diagnose strengths, gaps, and improvements

Why It Works

The model doesn’t actually need special command parsing — slash commands work because they are distinctive tokens that the system prompt has pre-loaded with meaning. When the user types /challenge [idea], the model retrieves the definition from its active context and executes accordingly. This is pure context engineering: you are giving the model a lookup table it carries into every conversation.

The behavioral brief and command library serve different functions:

  • The brief shapes how the model engages
  • The commands shape what it does on demand

Together they create a repeatable, personalizable interface that reduces the cognitive load of prompting.

For Coaches and Knowledge Entrepreneurs

This pattern is directly applicable to any professional who has recurring task types:

  • /brief [client name] — synthesize intake materials into a coaching brief
  • /reframe [client block] — offer three alternative framings of a stated problem
  • /assets [topic] — identify five content angles from a theme

The commands you define should map to the actual decisions and tasks you repeat most often. Once built, this becomes personal intellectual infrastructure — a set of cognitive shortcuts that compound over time.