Idea Vault

Idea Vault

by The Agent Ledger — The newsletter on building AI-native businesses.

Every solopreneur has more ideas than hours. The problem isn't the idea — it's that ideas live in your head, your notes app, a dozen open browser tabs, and the shower. They compete for attention equally, whether they're worth $0 or $100K/year.

The Idea Vault gives your ideas a home with a lifecycle. Capture them fast. Score them honestly. Review them systematically. Graduate the winners. Kill the rest with dignity.

What This Skill Does

• Captures new ideas in a structured format (quick or detailed)
• Scores ideas across 4 dimensions: revenue potential, effort, alignment, novelty
• Tracks ideas through a 5-stage pipeline: Raw → Researched → Validated → Active / Parked / Killed
• Reviews your backlog on a weekly quick-scan and monthly deep-review cadence
• Graduates validated ideas to project-tracker with a handoff brief
• Resurrects parked ideas on a schedule (some ideas are just early)
• Surfaces patterns across your idea history (what do you keep returning to?)

Setup

1. Create the Vault File

Create ideas/idea-vault.md in your workspace with this structure:

# Idea Vault

## Active Pipeline

### Raw (unscored captures)
<!-- New ideas land here first -->

### Researched (scored, needs validation)
<!-- Ideas that have been scored and deserve a closer look -->

### Validated (ready to build or launch)
<!-- Evidence gathered, worth committing to -->

## Parking Lot
<!-- Good ideas, wrong timing. Check quarterly. -->

## Archive
### Graduated → Active Projects
<!-- Ideas that became real projects -->

### Killed
<!-- Ideas you evaluated and consciously rejected. Record why. -->

2. Create the Scoring Reference

Save this in your working notes or at the top of idea-vault.md:

IDEA SCORE DIMENSIONS (rate each 1-5):
  R — Revenue Potential: Can this make meaningful money? (5 = $100K+ path)
  E — Effort Inverse: How easy to build/launch? (5 = hours, 1 = months)
  A — Alignment: Fits your skills, audience, and goals? (5 = perfect fit)
  N — Novelty: Is this defensibly different? (5 = clear gap in market)

Total = R + E + A + N (max 20)
  16-20: Prioritize immediately
  11-15: Research before committing
  6-10: Park with a review date
  1-5: Kill unless there's a specific reason to keep

3. Add AGENTS.md Standing Instructions (Optional)

Add to your AGENTS.md or HEARTBEAT.md:

## Idea Vault
- When I say "idea:" followed by anything, capture it to ideas/idea-vault.md under Raw
- During weekly review, flag any Raw ideas older than 14 days that haven't been scored
- During monthly review, check Parking Lot ideas with "revisit by" dates that have passed

Usage Patterns

1. Quick Capture

"Idea: subscription box for solopreneurs — curated AI tools + resources, $29/month"

The agent captures this to the Raw section with a timestamp and no further friction. Scoring happens later.

Capture format:

**[IDEA-###] Title**
Captured: YYYY-MM-DD
Source: shower / conversation / competitor / market gap / other
One-liner: [What it is in one sentence]
Status: Raw
Notes: [Optional initial context]

2. Score an Idea

"Score IDEA-042 — AI newsletter tools bundle"

The agent walks through each scoring dimension and fills in the rubric, producing:

**IDEA-042 Score**
R (Revenue): 4/5 — Newsletter tools have clear willingness to pay; bundle could do $500-2K MRR at scale
E (Effort): 3/5 — Curation work upfront, but no code needed; 2-4 weeks to MVP
A (Alignment): 5/5 — Perfect fit: AI tools, newsletter audience, solopreneur focus
N (Novelty): 3/5 — Bundle concept exists, but AI-specific + editorial curation is differentiator

Total: 15/20 — RESEARCH
Next step: Interview 5 newsletter subscribers about their AI tool spend. Set revisit: 2 weeks.

Move to Researched.

3. Research Update

"Update IDEA-042: talked to 8 newsletter readers, 6 said they'd pay $19-29/month for curated stack"

The agent logs the research finding, recalculates if score changes, and notes what validation step comes next.

4. Full Idea Review Session

"Run my idea review"

The agent goes through the pipeline systematically:

📦 IDEA VAULT REVIEW — [Date]

RAW (unscored): 6 ideas
  ⚠️ 2 ideas > 14 days old — flag for scoring or kill
  Oldest: IDEA-039 (23 days) — "AI-powered Etsy listing optimizer"
  
RESEARCHED: 4 ideas
  1 idea ready to validate (score 15+): IDEA-042
  2 ideas stalled (no activity 21+ days): IDEA-037, IDEA-040
  
VALIDATED: 1 idea
  IDEA-038: "Daily briefing cron template pack" — 18/20, waiting on capacity

PARKING LOT: 9 ideas
  3 ideas past revisit date — surfacing for review
  
KILLED (last 90 days): 12 ideas
  Pattern detected: 3 killed ideas involved physical products → possible blind spot

ACTION ITEMS:
  1. Score IDEA-039 or kill it — it's been sitting
  2. Advance IDEA-042 to Validated (validation evidence gathered)
  3. Review IDEA-037, IDEA-040 — stalled in Researched
  4. Revisit parked ideas #034, #031, #027 (past their review dates)

5. Graduate Idea to Project

"Graduate IDEA-042 to project-tracker"

The agent creates a handoff brief:

**IDEA-042 → PROJECT HANDOFF**
Idea: AI Newsletter Tools Bundle
Final Score: 16/20
Validation: 8 reader interviews, 75% positive; comparable product research done
Revenue Model: $19-29/month subscription, targeting 50 subs at launch
Launch Scope: Curation-only v1 (no code), biweekly drops, Google Doc delivery
Next Step: Outline curation criteria → select first 10 tools → write intro email

→ Create project file: projects/ai-newsletter-tools-bundle.md
→ Link to decision-log entry for the go/no-go decision
→ Archive IDEA-042 to Graduated section with graduation date

6. Kill an Idea Properly

"Kill IDEA-031 — I've looked at this 4 times and never start it"

**IDEA-031 KILLED — [Date]**
Reason: Repeatedly deprioritized; lack of genuine motivation despite good score
Lesson: High score ≠ high desire. Execution requires wanting to, not just calculating that I should.
Future signal: If this resurfaces in 6+ months, that's genuine pull — recapture then.

Killed ideas stay in the archive. The kill reasons are often the most valuable part of the vault.

7. Resurrect a Parked Idea

"Check parking lot — what's overdue for review?"

The agent surfaces parked ideas past their revisit date, with a brief context refresh and a binary prompt: keep parking, promote, or kill.

8. Idea Pattern Analysis

"What patterns do you see in my idea history?"

The agent reviews the vault (including killed ideas) and surfaces:

• What categories do you keep returning to? (content, tools, services, physical)
• What scores consistently high but never gets built? (motivation gap vs capacity gap)
• What kills most often? (market risk, effort, misalignment)
• Time between capture and kill/graduation (your natural evaluation pace)

Idea Pipeline Reference

Idea Score Quick Reference

R — Revenue Potential
  5: Clear $100K+/year path with reasonable effort
  4: $10K-100K/year plausible
  3: $1K-10K/year, modest ceiling
  2: Revenue possible but unclear how to monetize
  1: Essentially non-commercial

E — Effort Inverse (how EASY, not how hard)
  5: Can launch in hours or days (template, guide, curation)
  4: Launch in 1-2 weeks
  3: Launch in 1-2 months
  2: Launch in 3-6 months
  1: Year+ or requires co-founder/funding

A — Alignment (skills + audience + goals)
  5: You have the skills, it serves your audience, it fits your goals
  4: Strong in 2 of 3
  3: Strong in 1, workable in others
  2: You're stretching significantly
  1: Outside your wheelhouse entirely

N — Novelty / Differentiation
  5: Clear gap, no obvious competitor, strong point of view
  4: Crowded space but meaningfully differentiated
  3: Similar to existing solutions, defensible niche
  2: Similar to many competitors, differentiation unclear
  1: Undifferentiated commodity

Heartbeat Integration

Add to HEARTBEAT.md for periodic idea maintenance:

## Idea Vault Check (weekly)
- Flag any Raw ideas older than 14 days
- Surface Researched ideas with no activity in 21+ days
- Note any Parking Lot ideas past their revisit date
- Report: [# Raw] [# Researched] [# Validated] [# Parked]

For a monthly deep review, add a cron:

openclaw cron add \
  --name "monthly-idea-review" \
  --cron "0 10 1 * *" \
  --model "anthropic/claude-sonnet-4-6" \
  --session isolated \
  --message "Run a full Idea Vault review: score any unscored Raw ideas, flag stalled Researched ideas, surface Parking Lot items past their revisit dates, and identify any patterns in killed ideas from the past quarter. Report findings." \
  --announce \
  --to "[YOUR_TELEGRAM_CHAT_ID]" \
  --tz "America/Chicago"

Integration with Other Agent Ledger Skills

Customization

Adjust scoring weights: If revenue is your primary filter, weight R × 2 in your total. If you're optimizing for learning, weight A higher. The default equal-weight rubric is a starting point.

Add custom dimensions: Some builders add a 5th dimension: P (Personal Enjoyment) — would you actually enjoy building and running this? A score of 1 is a silent project killer.

Idea categories: Tag ideas with a category (product, service, content, tool, community, other) to spot patterns over time.

Team use: If you have a collaborator, add a "Submitted by" field and a "Champion" field (who owns evaluation). Ideas without a champion stall.

Kill threshold: Default is to kill ideas scored below 8. If you're early-stage and still exploring, lower to 6. If you're focused, raise to 12.

Troubleshooting

"My vault is full of Raw ideas and nothing gets scored" → Set a rule: no new captures without scoring one existing Raw idea first. Or batch-score every Sunday (15 minutes max).

"I score ideas high but never build them" → Add the P (Personal Enjoyment) dimension. A 4 REAM score that feels like work is a trap. Also check your goal-tracker — are ideas competing with committed goals?

"I keep resurrecting the same idea without making progress" → Treat the third resurface as a forcing function: either commit to a validation step with a deadline, or kill it permanently. Endless parking is just soft avoidance.

"My scores feel arbitrary" → Calibrate against your track record. Look at your graduated projects and their original scores. Adjust the rubric until past scores match what you now know about those projects.

Privacy Note

The Idea Vault stores everything locally. No content is transmitted externally unless you explicitly instruct your agent to share it. Your un-built ideas are competitively sensitive — treat this file accordingly (don't share it in group chats or multi-agent contexts with external access).

The Agent Ledger — Building AI-native businesses, one system at a time. Subscribe for weekly insights on agent setup, automation, and solopreneur strategy.

See also: references/advanced-patterns.md for idea tournament protocols, opportunity cost tracking, seasonal idea batches, and validation sprint frameworks.

License: CC-BY-NC-4.0 — Free to use and adapt for non-commercial purposes. Please credit The Agent Ledger.