prompt-catalog

Reference: Prompt Catalog

Primary Prompts

• author-specification: Interactive spec authoring; reads org presets during discovery only.
  • Next Steps: derive-implementation-plan
  • Model Guidance: Use a reasoning-capable model; temperature 0.2–0.4; allow tool use to scan repo.
• derive-implementation-plan: Builds executable plan from approved spec; ignores presets.
  • Next Steps: decompose-feature-work
  • Model Guidance: Reasoning-capable; temperature 0.2–0.3; emphasize structure/traceability.
• decompose-feature-work: Creates feature breakdown with WIs, tests, docs; ignores presets.
  • Next Steps: implement-next-work-item
  • Model Guidance: Balanced coder model; temperature 0.2–0.4; prefer deterministic outputs.
• implement-next-work-item: Implements a selected WI; ignores presets.
  • Next Steps: author-adr if deviations; otherwise pick next WI
  • Model Guidance: Strong code generation + tool use; temperature 0.1–0.2; minimal diffs.
• author-adr: Records architecture decisions.
  • Next Steps: propagate-adr
  • Model Guidance: Analytical model; temperature 0.2–0.3; require explicit impacts and alternatives.
• propagate-adr: Applies accepted ADRs across artifacts.
  • Next Steps: revisit spec/plan if required; then continue WI implementation
  • Model Guidance: Structured editing; temperature 0.1–0.2; prioritize consistency checks.
• author-refactor-plan: Plans refactors aligned with spec.
  • Next Steps: implement-next-work-item or derive follow-up tasks
  • Model Guidance: Analytical; temperature 0.2–0.3; list risks, tests, and rollback.

Inputs Matrix (high level)

Notes

• Presets are suggestion-only during author-specification discovery and are ignored by all other prompts.
• Every prompt enforces traceability using ID prefixes (see id-conventions.md).
• For existing projects, prompts analyze current code and extend patterns rather than replace them.