deepbook-research— 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 deepbook-research idea completed
# Or use "failed" / "aborted" if it ended that way.
command -v suiperpower >/dev/null 2>&1 && suiperpower track deepbook-research 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

Pulls DeepBook on-chain trading data (pools, recent trades, depth, fees) and turns it into a list of product opportunities. The skill identifies underserved pairs, low-spread niches, and observable gaps (no aggregator coverage, no charting tool, no MEV resistance, no specific market-making strategy). Output is a ranked list of candidate ideas grounded in real DeepBook activity, not speculation.

When to use it

• The user wants to build on DeepBook but is not sure what.
• The user is mid-idea-validation and wants to confirm DeepBook traction.
• The user is sponsor-track-aligned (Sui Overflow DeepBook track) and needs a load-bearing integration angle.

When NOT to use it

• The user wants to build a DeepBook integration with a chosen idea, route to deepbook-orderbook.
• The user wants general Sui idea search, route to find-next-sui-idea.
• The user wants Walrus-specific research, route to walrus-research.

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

Inputs

• The user's interest in DeepBook (curious, picking an idea, validating).
• Optional: a specific pair the user is interested in (e.g. SUI/USDC, DEEP/USDC).
• Optional: the user's chain experience (helps tailor the analysis depth).

Outputs

A research block written to .suiperpower/idea-context.md (or a new .suiperpower/research-deepbook-<timestamp>.md if no idea is chosen yet):

## DeepBook research, <timestamp>

### Pools surveyed
- <pair>: <observed depth>, <observed daily volume>, <fee tier>, <maker concentration>

### Observable gaps
1. <gap>: <evidence from data>, <product idea this enables>
2. ...

### Underserved pairs
- <pair>: <why underserved>, <product idea this enables>

### Risks
- <risk to building on DeepBook today>: <mitigation>

### Citations
- <DeepBook docs link, RPC query examples, dashboards>

Workflow

1. Confirm scope

• Is the user open to any DeepBook angle, or focused on a specific pair or product type?

1. Survey pools

• List active DeepBook v3 pools (read-only RPC, or via DeepBook indexer if available).
• Note for each: trading pair, fee tier (basis points), 24h volume, current depth at +/-1% of mid.
• Identify the top 5 by volume; the top 5 by depth; the bottom 5 by spread efficiency.

1. Walk the gap categories

• Aggregator coverage: which pools are indexed by aggregators (Cetus, Aftermath, Bluefin, Hop, etc.)? Pools that are active but not aggregated are an opportunity.
• Charting / data tooling: is there a public dashboard for DeepBook activity beyond the official one? If thin, opportunity.
• Market-making bots: are there public bot frameworks? If thin, opportunity for a Sui-native MM strategy as a product.
• MEV resistance: is there observable MEV in the order flow? If yes, opportunity for a private-RFQ or bundled-PTB product.
• Specific pairs: which pairs have no liquidity, no tooling, or no integration with consumer-facing apps?

1. Surface underserved pairs

• Stablecoins paired with sponsor tokens (DEEP, WAL, SCA): often thin liquidity, high spread.
• Long-tail tokens with active community but no DeepBook listing.
• Cross-asset pairs that depend on bridged tokens (look for bridge-native pairs).

1. Identify risks

• DeepBook v3 may have specific listing or fee constraints; document.
• Aggregator dominance: if 90% of volume routes through one aggregator, the moat for new aggregators is small.
• User audience: most Sui DeepBook users are sophisticated; consumer-facing products need a different audience hypothesis.

1. Cite the work

• Every claim is tied to a DeepBook RPC query, an explorer link, an aggregator docs page, or a dashboard.
• Refuse to make claims without citations. "I assume" is not citation.

1. Writeback

• Append to the chosen output file.

Quality gate (anti-slop)

Before reporting done:

• Is every quantitative claim (volume, depth, fee) tied to a citation (RPC query result, dashboard link, docs)?
• Are the gaps named with specific evidence, not abstract "DeepBook needs X"?
• Did the analysis include at least one risk, not just upside?
• Did the writeback happen?
• Did the analysis avoid recommending a product the candidate cannot ship in their stated timeline?

If any answer is no, the skill keeps working.

References

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

• references/deepbook-data-queries.md: RPC query patterns and indexer endpoints for DeepBook.
• references/gap-categories.md: The categories of gaps to walk through.

Knowledge docs:

• skills/data/sui-knowledge/sponsor-docs/deepbook.md: DeepBook integration knowledge doc.
• skills/data/sui-knowledge/04-protocols-and-sdks.md: SDK and integration overview.

Use in your agent

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

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