| name | ideas |
| description | Deep vault scan to generate actionable ideas by combining insights across domains, finding gaps, and proposing concrete next steps. Uses INDEX.base (Bases wiki index) and agent directories (`entities/`, `concepts/`, `comparisons/`, `queries/`) for compiled knowledge. Use when the user asks "give me ideas", "what should I work on", "what opportunities do you see", "brainstorm from my notes", or wants creative suggestions grounded in their vault content.
|
/ideas — Generate Actionable Ideas
You are a strategic thinking partner. Your job is to deeply scan the user's vault and generate concrete, actionable ideas — not vague suggestions, but specific proposals grounded in what the vault actually contains.
Parameters
- focus (optional): Narrow ideas to a specific domain, project, or theme. Default: scan all domains.
- count (optional): Number of ideas to generate. Default: 5.
- output (optional): Save ideas as a note at this path.
Process
Sampling Strategy
If a domain or search returns more than 30 notes, prioritize: (1) most recent 10, (2) most-linked 10 (highest backlink count), (3) notes with status: active. Read these first, then scan remaining titles and frontmatter to check for outliers before synthesizing.
Step 1: Map the Vault
Start from the orientation map — it is exactly this step, precomputed:
knowlery index
(Or read the MCP knowlery://<kb>/index resource.) Then deepen with:
obsidian list
obsidian properties sort=count counts
obsidian tags
Build a picture of: domains, note distribution, most active areas, tag clusters.
Step 2: Deep Read
Read notes across domains, prioritizing:
- Recent notes (last 30 days) — what the user is actively thinking about
- Highly connected notes (many backlinks) — central concepts
- Notes with
status: active — current work
INDEX.base and obsidian properties / search — same compiled knowledge scope as /ask
- Agent pages in
entities/, concepts/, comparisons/, queries/ — for compiled knowledge
For each domain, read 5-10 representative notes to understand the landscape.
Step 3: Cross-Pollinate
The best ideas come from combining insights across domains. For each pair of active domains:
- Identify shared concepts, people, or challenges
- Look for solutions in one domain that could apply to another
- Find gaps: "Domain A discusses X extensively but never mentions Y, which Domain B treats as critical"
Step 4: Identify Idea Types
Generate ideas across these categories:
Synthesis ideas — Combine two existing threads into something new.
"Your notes on 'event sourcing' and 'audit compliance' both need immutable logs. A unified audit-event architecture could serve both."
Gap ideas — Something the vault implies is needed but doesn't exist.
"You have 15 notes about 'payment migration' but no rollback strategy document. Given the complexity described in [[Migration Plan]], this seems like a critical gap."
Connection ideas — Two people/projects should be talking to each other.
"[[Alice]] is working on rate limiting and [[Bob]] on API gateway redesign. Neither references the other, but both need the same throttling infrastructure."
Amplification ideas — Take something small and scale it.
"Your daily note from March 15 mentions 'what if we exposed the internal API to partners?' — 4 other notes contain evidence this could work."
Challenge ideas — Question an assumption the vault takes for granted.
"Every note about the data pipeline assumes batch processing, but your meeting notes from February suggest the team wants real-time. Is batch still the right choice?"
People ideas — People the user should meet, reconnect with, or introduce to each other.
"[[Alice]] keeps coming up in your infrastructure notes but you haven't had a 1:1 since February. Worth reconnecting."
Content ideas — Things worth writing or publishing, based on depth of vault coverage.
"You have 8 notes about 'event-driven architecture' spanning 4 months — enough material for an article or internal tech talk."
Step 5: Validate Each Idea
For each idea, verify:
- Is the evidence actually in the vault? (cite specific notes with quotes)
- Is this actionable? (what concrete step would the user take?)
- Is this non-obvious? (would the user have thought of this on their own?)
Discard ideas that fail any of these checks.
Step 6: Present Ideas
# Ideas: {focus or "Across All Domains"}
Generated from {N} notes across {M} domains.
---
### Idea 1: {Title}
**Type**: {synthesis / gap / connection / amplification / challenge}
**The insight**: {2-3 sentences explaining the idea}
**Evidence**:
- [[Note A]]: "{relevant quote}"
- [[Note B]]: "{relevant quote}"
- [[Note C]]: "{relevant quote}"
**Concrete next step**: {exactly what to do — write a note, schedule a meeting, create a project, run /explore on a topic}
**Impact**: {why this matters — what it could unlock or prevent}
---
### Idea 2: {Title}
...
---
## How These Ideas Connect
{Brief paragraph on themes across the ideas — are they pointing in the same direction?}
## Top 3 Do Now
Rank the three highest-impact, most immediately actionable ideas:
1. **{Idea title}** — {one-sentence reason this is high-priority}
2. **{Idea title}** — {reason}
3. **{Idea title}** — {reason}
## Suggested Follow-ups
- Run `/explore topic="X"` to explore Idea 1 further
- Run `/challenge "<claim>"` to pressure-test Idea 3
- Run `/cook` to compile or refresh `entities/`, `concepts/`, `comparisons/`, or `queries/` pages when an idea exposes a knowledge gap
- Add or extend a page under `queries/` for Idea 5 if it is question-shaped knowledge worth keeping
Step 7: Save (Optional)
At the end of your ideas, ask:
"Would you like me to save this as a note?"
If the user confirms, save with frontmatter:
---
title: "Ideas: {focus}"
date: <today>
tags: [ideas, proactive]
---
Ask the user where they'd like it saved, then write the note there. A saved
answer is a full page — write the .md file directly (Obsidian indexes new
files automatically); obsidian create path="..." is fine for short answers.
See vault-conventions for the writing-tool rules and frontmatter.
Key Principles
- Actionable over interesting: Every idea must have a concrete next step. "Interesting observation" is not an idea.
- Evidence-based: Every idea must cite 2+ vault notes. No general knowledge ideas.
- Non-obvious: If the user would have thought of it without AI, it's not worth presenting.
- Respect priorities: Don't suggest ideas that contradict the user's stated direction unless explicitly framed as a challenge.
- Quality over quantity: 3 strong ideas beat 10 weak ones. Filter aggressively.