Ideas Folder Documentation
The Exploration Space
Before work becomes scoped — with a clear desc.md, a step-by-step plan, and a critic — it starts as an idea. The Ideas folder is where broader explorations, unbounded thinking, and not-yet-shaped possibilities live.
The distinction between ideas/ and scoped/ is about readiness, not importance:
scoped/= “I know what to change” — boundaries are drawn, deliverable is definedideas/= “I’m exploring something bigger” — scope isn’t fully defined yet, might need sensemaking before a plan
The Ideas Folder Structure
devdocs/ideas/
langfuse_adaptation/
desc.md
sensemaking.md
scorer_model_idea.md
event_driven_architecture/
elaboration.md
roadmap.md
future/
parallel_filler_subagents/
desc.md
Ideas are named, not numbered. Names are descriptive — they tell you what the idea is about at a glance. Scoped items get numbers because they’re in a sequential pipeline. Ideas aren’t in a pipeline yet.
What Goes Here
- Feature explorations that don’t have clear boundaries yet
- Architecture possibilities being evaluated
- Sensemaking analyses for complex decisions
- Elaborations of vague requirements
- Roadmaps for multi-step visions
- Anything that needs thinking before it needs planning
How Ideas Differ from Scoped Work
| Ideas | Scoped | |
|---|---|---|
| Readiness | Still being shaped | Ready to implement |
| Naming | Descriptive names | Sequential numbers |
| Key docs | elaboration.md, sensemaking.md, desc.md (exploratory) | desc.md (specific), step_by_step_impl_plan.md, critic.md |
| Primary commands | /elaborate, /sense-making, /roadmap | /task-desc, /task-plan, /critic |
| Scope | May be unclear, multi-faceted | Clear boundaries, defined deliverable |
| Structure | Flexible — whatever the idea needs | Standardized — desc, plan, critic, test |
When an Idea Matures
When an idea becomes clear enough to implement, it moves to scoped/:
mv devdocs/ideas/langfuse_adaptation/ devdocs/scoped/feat_19_langfuse_adaptation/
It gets a number and enters the implementation pipeline. At that point, the exploratory docs (sensemaking, elaboration) become context for creating the structured docs (desc.md with success criteria, step_by_step_impl_plan.md, critic.md).
Not all ideas mature. Some stay as ideas indefinitely — valuable context for future decisions but never implemented directly. Some get archived when the direction changes. Both are fine.
The future/ Subfolder
Ideas that are planned but not actively being explored go in ideas/future/:
devdocs/ideas/future/parallel_filler_subagents/
This keeps them visible (the AI sees them when reading the ideas folder) without cluttering the active exploration space. When you’re ready to explore, move them up:
mv devdocs/ideas/future/parallel_filler_subagents/ devdocs/ideas/parallel_filler_subagents/
Best Practices
- Start messy — a single markdown file with raw thoughts is a valid idea. Structure it later if it survives.
- Use sensemaking for complex ideas — if the idea is ambiguous or multi-faceted, run
/sense-makingon it before trying to scope it. - Don’t rush to scoped/ — an idea moved to scoped/ too early produces a weak plan. Let ideas breathe.
- Link related ideas — if two ideas interact, reference each other. This helps when one matures and you need to consider the other.
- Archive, don’t delete — abandoned ideas go to
archive/, not the trash. The reasoning behind abandoned ideas prevents re-exploring dead ends.