validate-idea— agent skill

Preamble (run first)

# Suiperpower telemetry. Routes through the CLI so projects.json + Convex stay in sync.
# Silent on failure so the skill never blocks. Opt out: SUIPERPOWER_TELEMETRY=off.
#
# AGENT NOTE: when this skill finishes, run the matching completion command:
#   suiperpower track validate-idea idea completed
# Or use "failed" / "aborted" if it ended that way.
command -v suiperpower >/dev/null 2>&1 && suiperpower track validate-idea idea started >/dev/null 2>&1 &
true

If TEL_PROMPTED is no, before doing real work, ask the user:

Help suiperpower get better. We track which skills get used and how long they take. No code, no file paths, no PII. Change anytime in ~/.suiperpower/config.json.

>

A) Sure, anonymous B) No thanks

Write the answer to ~/.suiperpower/config.json telemetryTier field and create ~/.suiperpower/.telemetry-prompted. Then continue.

What this skill does

Pressure tests a Sui product idea on four dimensions, demand evidence, competition, feasibility (real research-grade unknowns only, not "this sounds like a lot of code"), and Sui-native fit, and produces a written go/no-go recommendation. The skill is designed to kill weak ideas early. A "no go" with a clear reason is more valuable than a "go" with hand-waving. With AI-assisted pacing, do not kill ideas on apparent scope alone, application-layer scope is rarely the bottleneck.

When to use it

• After find-next-sui-idea has written a chosen idea to .suiperpower/idea-context.md and the user wants a second pass before scaffolding.
• The user has an idea (their own, not from the corpus) and wants pressure testing.
• The user is mid-build and questioning whether to continue. Late validation is still useful.

When NOT to use it

• The user has not picked an idea yet. Route to find-next-sui-idea.
• The user wants competitive landscape only. Route to competitive-landscape.
• The user wants a brutal critique of the existing product. Route to roast-my-product.

If you activated this and the user actually wants something else, consult skills/SKILL_ROUTER.md and hand off.

Inputs

• .suiperpower/idea-context.md if it exists.
• The idea in one or two sentences if no context file.
• Optional: any prior validation work the user has done (interviews, surveys, search-trend data).

Outputs

A validation block appended to .suiperpower/idea-context.md:

### Validation, <timestamp>
- demand evidence: <strong | medium | weak>, with one to three citations
- competition: <none | one | crowded>, with named competitors (Sui-specific and adjacent-chain)
- feasibility: <yes | research-blocked> (research-blocked means the idea requires novel cryptographic, consensus, or L1-level work the user must drive personally, not just "lots of code"; AI-assisted pace handles application-layer scope)
- Sui-native fit: <strong | medium | weak>
- recommendation: <go | go-with-pivot | no-go>
- reasoning: <one paragraph>
- next experiments: <bullet list of 1-3 cheap tests to run before deeper investment>

The recommendation is one of go, go-with-pivot, no-go. go-with-pivot means the underlying market is real but the proposed product needs a specific change.

Workflow

1. Read the context

• Open .suiperpower/idea-context.md.
• If missing, ask the user to describe the idea in one or two sentences and write a minimal context file before proceeding.

1. Walk demand evidence

• Has anyone paid for something similar (on Sui, on EVM, on Solana, off-chain)? Cite at least one.
• Are there public discussions, fundraises, or sponsor priorities pointing at the category?
• What is the falsifiable signal that demand is real? (Search trends, paying customers elsewhere, sponsor RFP, repeated user request.)

1. Walk competition

• List Sui-native competitors. None? One? Crowded?
• List adjacent-chain competitors. What is their state (live, growing, struggling, dead)?
• For each, what is the candidate's specific differentiation in one sentence?

1. Walk feasibility

• Identify any genuinely research-grade dependency (novel cryptography, custom consensus, untested L1 work, brand-new oracle infra the user must build). Application-layer scope, Move modules, sponsor SDK integrations, and frontends are not research-grade, AI-assisted pace handles them.
• If a hard external deadline exists (Overflow cutoff, grant milestone), flag any research-grade dependency as a hard feasibility risk against that deadline. Otherwise, mark feasibility yes.
• Do not invent "cuts to scope" unless the user asks for the minimum surface. The agent can usually build the full thing.

1. Walk Sui-native fit

• What primitive does the idea depend on (Object model, Walrus, DeepBook, Kiosk, zkLogin, parallel execution, low fees)?
• What does Sui give the idea that EVM or Solana would not?
• "Sui has lower fees" alone is not a moat.

1. Compose the recommendation

• go if at least three of four dimensions are strong or medium and no dimension is weak.
• go-with-pivot if one dimension is weak but the underlying category is healthy.
• no-go if two or more dimensions are weak, or if the idea is on the non-sui-native rejection list with no override.

1. Suggest cheap next experiments

• 1 to 3 tests, each runnable in under a day, that would move at least one weak dimension to medium.
• Examples: post the idea on a Sui community channel and count engagement; run a 10-person user interview; build a clickable prototype.

1. Writeback

• Append to .suiperpower/idea-context.md.

Quality gate (anti-slop)

Before reporting done:

• Is the demand-evidence claim backed by at least one citation that is not "founder intuition"?
• Is the competition list real, with named projects, not abstract "competitors exist"?
• Is feasibility called out only on real research-grade unknowns, not on "this sounds big"?
• Did the recommendation actually land on go, go-with-pivot, or no-go, not a hedge?
• Did the writeback happen?

If any answer is no, the skill keeps working.

References

On-demand references (load when relevant to the user's question):

• references/demand-evidence-questions.md: Specific questions to test demand claims.
• references/no-go-criteria.md: Patterns that should trigger a no-go regardless of enthusiasm.

Knowledge docs:

• skills/data/ideas/sui-native-gaps.json: Curated gap list, useful for pivot suggestions.
• skills/data/sui-knowledge/06-opensource-research.md: Where to look for adjacent research and existing competitors.

Use in your agent

• Claude Code: claude "/suiper:validate-idea <your message>"
• Codex: codex "/validate-idea <your message>"
• Grok Build: run grok, then /validate-idea <your message> in the session
• Cursor: paste a chat message that includes a phrase like "validate my idea", or load ~/.cursor/rules/validate-idea.mdc and reference it.

If you activated this and the user actually wants something else, consult skills/SKILL_ROUTER.md and hand off.