The Prompt Bank

What This Is

Real prompts from ~6 months of K-NOME development — the gardener’s shorthand, not polished templates. These are the phrases that move the ecosystem forward, organized by intent.

1. Comprehensive Audit

The audit prompt is the starting point for any session:

“Review this crate for deep debt, evolution gaps, hardcoded values, unsafe code that could be safe, mocks in production, and overstep cleanup. Proceed with fixes.”

Variations target specific concerns:

“Analyze external dependencies and evolve to Rust.” “Large files (>800L) should be refactored smart rather than just split.”

2. Constraint Evolution

Prompts that apply specific constraints:

“Evolve this to accept a branch parameter instead of hardcoding main.” “Replace .expect() with Result-based error handling in production code.” “Centralize this magic number as a named constant.” “Make this agnostic — primal code only has self-knowledge.”

3. Cross-System Coordination

Moving patterns between springs and primals:

“Apply the same pattern from hotSpring to wetSpring.” “Propagate this fix across all crates that use the same idiom.” “Cascade from VPS and review.”

4. Deep Debt

Targeted debt elimination:

“Proceed to execute on all remaining deep debt and evolution gaps.” “Mocks should be isolated to testing; any in production should be evolved to complete implementations.” “Hardcoding should be evolved to agnostic and capability-based.”

5. Handoff and Documentation

Generating documentation and coordination artifacts:

“Update root documentation — CONTEXT.md, EVOLUTION_QUEUE.md, CHANGELOG, test counts.” “Write a wateringHole handoff for this wave.”

6. The Proceed Prompt

The simplest and most powerful K-NOME prompt:

“Proceed.”

This prompt trusts the AI to continue the current work without additional direction. It works because the conversation context carries the intent — the AI knows what was audited, what was fixed, what remains.

The proceed prompt is the mentoring pattern in miniature: the gardener has set the direction, the AI tends the growth, and “proceed” means “keep going, I trust the trajectory.”

Pattern Observations

• Near-duplicates reflect reuse — the same audit prompt appears in many springs because the same patterns recur
• Constraint language evolves — early prompts say “fix this”; mature prompts say “evolve this to modern idiomatic Rust”
• Garden language is natural — “tend,” “propagate,” “fossil,” “prune” emerge organically from the biological framing
• Geological language scales up — “stadial gate,” “interstadial,” “glacial goal” describe ecosystem-level K-NOME

The prompt bank is not a template library. It is a vocabulary — a living record of how one gardener talks to the garden. Your vocabulary will be different. The constraint is the same: the conversation is the interface.