Think

Route $ARGUMENTS to the right thinking mode based on intent. Detect the mode from context — do NOT ask.

Routing

If intent is ambiguous, default to Deep Analysis.

Creative

Divergent ideation — quantity and variety over polish.

1. Restate the challenge in one sentence
2. Obvious solutions — 2-3 conventional approaches (acknowledge, then move past)
3. Wild ideas — 5-7 unconventional approaches. Mix:
   • Inversion (what if we did the opposite?)
   • Analogy (how does a different domain solve this?)
   • Removal (what if we deleted the constraint?)
   • Combination (what if we merged two approaches?)
4. Diamond pick — which 1-2 wild ideas have real potential and why
5. Next step — one concrete action to explore the best idea

Deep Analysis

Multi-angle exploration for complex topics.

1. Frame the question precisely — what is actually being asked?
2. Technical — mechanics, constraints, and trade-offs
3. Practical — what does this look like in practice? What's the effort and friction?
4. Strategic — how does this fit the bigger picture? What does it enable or block?
5. Tensions — where do the angles disagree? What can't be optimized simultaneously?
6. Synthesis — what the analysis reveals that wasn't obvious at the surface

Examples

Example 1 → First Principles

User: "why is our test suite so slow?"
→ Intent: root cause, decompose
→ Invoke first-principles: "Why is the test suite slow?"
→ Break down: what makes tests slow? → I/O, parallelism, fixtures, network calls
→ Challenge each assumption: are all these tests necessary? Is the slowness uniform?

Example 2 → Council

User: "should we use a monorepo or separate repos?"
→ Intent: debate, weigh options
→ Invoke council with the question
→ Multiple perspectives argue in parallel, then synthesize

Example 3 → Creative

User: "how could we make onboarding more engaging?"
→ Intent: brainstorm, possibilities
→ Restate: "How do we make first-run onboarding feel less like a form?"
→ Obvious: wizard steps, progress bar, skip option
→ Wild: game mechanic for first task, video from the AI, co-pilot mode for first session
→ Diamond pick: co-pilot mode — high novelty, directly addresses the friction

Example 4 → Deep Analysis

User: "what are the tradeoffs of server-side rendering?"
→ Intent: explore tradeoffs, implications
→ Technical: hydration cost, TTFB vs TTI, caching strategies
→ Practical: team expertise, deployment complexity, SEO requirements
→ Strategic: enables progressive enhancement, constrains client-side interactivity
→ Tensions: performance vs interactivity, simplicity vs capability
→ Synthesis: SSR is a distribution of complexity, not a removal of it

Anti-patterns

• Don't ask which mode to use. Detect from context and commit. The routing table is the decision — use it.
• Don't use Deep Analysis as the default for everything. "Should I do X or Y?" is Council. "Why is X broken?" is First Principles. Reserve Deep Analysis for genuine multi-angle exploration.
• Don't produce shallow Creative output. Wild ideas must genuinely challenge constraints — listing 5 obvious variations is not creative thinking.
• Don't skip the Tensions section in Deep Analysis. If you can't find a tension, you haven't analyzed deeply enough.
• Don't combine modes in one response. Pick one and commit. Mixing modes produces unfocused output.
• Don't over-explain the routing. One line is enough: "This is a root-cause question — First Principles." Then do the work.

Rules

• Mode detection is mandatory — classify before starting, state it in one line
• First Principles and Council are delegated — always invoke those skills rather than reimplementing them
• Creative and Deep Analysis run inline — follow the steps above directly
• Tensions are non-negotiable in Deep Analysis — name at least one real conflict between the angles
• Wild ideas in Creative must be genuinely unconventional — if a PM would have thought of it in 30 seconds, it's not wild enough