feat: onchain random idea picker, dual-state search & streamlined publish flow

[y4motion]

Summary

Enhances the ideation skill with onchain knowledge graph integration, a random idea picker mode, and a streamlined publishing workflow.

Addresses Mission 03C — Skill Enhancement from the Intuition Missions board.

What's Changed

🎲 Random Idea Picker (Mode B)

• New entry point: users can say "random idea", "inspire me", or "what should I build"
• Queries the Intuition mainnet GraphQL API for ideas published as atoms with the [Idea] → [top project ideas for] → [Intuition] triple pattern
• Shows community backing (staker count, $TRUST staked)
• Falls back to GitHub ideas repo if onchain list is empty
• Supports re-rolling (up to 5 times) or switching to full ideation mode

🔍 Dual-State Checking (Step 1b)

• Check 1: Onchain Knowledge Graph — searches atoms by keyword, ordered by vault shares
• Check 2: GitHub Ideas Repository — searches issues/PRs
• Check 3: Onchain Ideas List — checks if idea already exists in the curated list
• Prevents duplicate work and surfaces existing community interest
• Clear guidance for each scenario (exists onchain, exists on GitHub only, fresh)

💰 Streamlined Publish Flow (Step 5)

• Pre-flight checks — queries atom cost, triple cost, curve ID, and checks for existing atoms BEFORE any user interaction
• Cost transparency — shows exact $TRUST costs upfront with clear summary
• Duplicate prevention — detects existing atoms and reuses them instead of failing
• Batch triple creation — creates all relationship triples efficiently
• Deposit preview — shows exact shares the user will receive before executing
• Ideas list linking — auto-creates [Idea] → [top project ideas for] → [Intuition] triple for discoverability

🛡️ Enhanced Error Handling

• IPFS pin failures (with fallback)
• Insufficient $TRUST balance (with guidance)
• Atom already exists (reuse instead of fail)
• Transaction reverts (mapped to specific causes)
• Random Picker returning no results (GitHub fallback)
• Idea remixing from Random Picker (create new, reference original)

Files Changed

• SKILL.md — Main skill definition (350 → 430+ lines)
• README.md — Updated with v2 feature documentation

Testing

• GraphQL queries validated against https://mainnet.intuition.sh/v1/graphql
• Trigger phrases tested for both Mode A and Mode B entry points
• Error paths documented with specific recovery instructions

Closes #1

[y4motion]

Hi @jeremie-olivier and @danielamodu 👋

This PR adds three enhancements to the ideation skill:

1. Onchain Random Idea Picker — queries the Intuition GraphQL API to pick a random atom from existing onchain ideas, so creators get fresh inspiration without repeating already-posted concepts
2. Dual-state checking — validates both GitHub issues and onchain atoms before suggesting ideas, preventing duplicates across both layers
3. Streamlined Step 5 publish flow — adds pre-flight checks, batch triple creation previews, and deposit amount estimates before committing onchain

Would love your feedback when you get a chance!

[danielamodu]

Hey! Thanks for the tag — this looks great. The onchain random idea picker is exactly the kind of friction-reducer the ideation flow needed. A few thoughts:

1. The dual-state checking (GitHub + onchain) is smart — prevents the awkward situation where someone proposes an idea that's already been atomized onchain but not tracked on GitHub.

2. For the batch triple creation preview in Step 5 — does it show estimated gas/deposit costs before committing? That would help contributors gauge onchain costs upfront.

3. One thing worth considering: could the random idea picker weight by "staleness" — surfacing atoms that haven't had activity in X days first? Helps resurface good ideas that got buried.

Happy to test this and give more detailed feedback once I've had a chance to run through it.

[y4motion]

Thanks @danielamodu — really glad this resonates! Great questions, and I went ahead and shipped all three:

1. Gas/deposit costs in preview — already implemented ✅

Step 5a (Pre-flight checks) queries getAtomCost() and getTripleCost() from the MultiVault contract and presents a full breakdown before any transaction:

• Atom creation cost
• Triple creation costs (up to 4)
• User's stake amount
• Estimated total in $TRUST

No surprises — the user confirms before any tx fires.

2. Staleness weighting — just shipped in 1e574fe 🚀

Loved this idea so much I built it. The Random Idea Picker now has three discovery modes:

Mode	Strategy	How it works
🔥 Popular	Top backed	Sort by totalShares desc, pick from top 10
💎 Hidden Gems	Staleness-weighted	Filter: positionCount > 0 AND updatedAt > 14 days ago — ideas with believers but buried
📈 Rising	Momentum scoring	stakers / days_since_last_activity — identifies ideas gaining traction fastest

Users choose their mode or say "surprise me" for pure random. They can switch modes mid-session too.

Each mode has a tailored presentation — Hidden Gems says "This idea had believers but got buried. Could you be the one to revive it?" while Rising shows the momentum score.

3. Testing

Quickest way:

npx skills add intuition-box/intuition-ideation-skill

Then try: "inspire me", "show me hidden gems", or "what's trending" to trigger Mode B with different discovery strategies. GraphQL queries hit mainnet live.

Would love your feedback on the discovery modes — especially whether the 14-day dormancy threshold for Hidden Gems feels right, or if you'd prefer it configurable.

[danielamodu]

This is excellent — the three discovery modes are exactly the right abstraction. A few thoughts after testing:

Hidden Gems threshold: 14 days feels right as a default but configurable would be better long-term. Some ideas are seasonal (e.g. infrastructure ideas that only matter during certain protocol phases). A slider or parameter like dormancy_days would let power users tune it.

Rising mode: the stakers / days_since_last_activity formula is clean. One edge case — new ideas (< 3 days old) will always score high on momentum even with 1 staker. Worth adding a minimum age filter like created_at > 3 days ago to avoid surfacing brand new ideas as "rising."

Surprise me: love it. This is the one I'd use most — reduces selection paralysis.

Gas preview: the pre-flight breakdown is exactly what was missing. No notes here, this is perfect.

Overall this feels production-ready. The mode switching mid-session is a nice touch — keeps exploration fluid without forcing a restart.

[y4motion]

Shipped both in 94390f1 — plus a few extras:

1. Configurable dormancy_days ✅

Hidden Gems threshold is now a tunable parameter (default: 14). The skill proactively asks:

"The default dormancy threshold is 14 days. Want to adjust? Some seasonal infrastructure ideas may need 30+ days. Just say a number or 'default'."

This handles exactly your point about seasonal/infrastructure ideas.

2. Rising mode minimum age filter ✅

New min_rising_age_days parameter (default: 3). Ideas younger than 3 days are excluded from momentum ranking — prevents the 1-staker false positive you flagged.

Includes a graceful fallback: if ALL ideas are < 3 days old, the filter is bypassed so results are never empty.

3. Selection reasoning (bonus) 🧠

Every picked idea now outputs SELECTION_REASON — a human-readable explanation of WHY it was chosen:

• Popular: "Top-backed idea with 15 stakers and 2400 $TRUST staked"
• Hidden Gems: "Had 8 believers but dormant for 23 days — ripe for revival"
• Rising: "Gaining 2.45 stakers/day — fastest growing in the pool"

4. Threshold transparency (bonus) 📊

Presentations now show the active thresholds (DORMANCY_THRESHOLD, MIN_RISING_AGE) so users know exactly what parameters drove the selection.

Parameter	Default	User-tunable
dormancy_days	14	✅
min_rising_age_days	3	✅

README updated with v2.2 changelog. Ready for another round of feedback if anything else stands out.