Imported from bankrbot/skills (delu-oracle/SKILL.md). Install upstream with npx skills add bankrbot/skills --skill delu-oracle. Copyright stays with the author.

delu-oracle

Intelligence layer for any Base trading agent. Pass one Base EVM contract address (or cashtag like $BNKR) and get back a flat decision header an agent can act on in a single hop — action, conviction, entry/stop/size, and a one-line delu-voiced read — with the full cognition report underneath for the why.

Scout, auditor, and quant run server-side on every call. The full observed block is always present. Social signal (checkr) is opt-in via ?social=true.

Endpoint

GET https://x402.bankr.bot/0xed2ceca9de162c4f2337d7c1ab44ee9c427709da/delu-oracle/analyze/{ca}

Base is the only supported chain — no chain parameter needed.

Parameters

⚠️ Sequential calls required

Do not call this endpoint in parallel. The x402 upto scheme uses a single-use Permit2 signature per authorization. Parallel calls with the same payer wallet result in 402 Payment could not be verified on all but the first.

Always call sequentially — one CA at a time, await the full response, then call the next.

// ✅ correct
for (const ca of watchlist) {
  const result = await oracle.analyze(ca);
  process(result);
}

// ❌ wrong
const results = await Promise.all(watchlist.map(ca => oracle.analyze(ca)));

The decision header — read this first

Flat, no traversal needed:

"decision": {
  "action": "ENTER",       // ENTER | WATCH | AVOID
  "conviction": 71,        // 0-100
  "direction": "long",
  "entry_low": 0.00051,
  "entry_high": 0.00053,
  "stop": 0.00048,
  "size_pct": 3.1,
  "read": "one line, delu voice"
}

Simple gate: decision.action === "ENTER" && decision.conviction >= 70 && confidence >= 0.6

action maps from verdict: strong_buy/buy → ENTER, hold → WATCH, avoid/drop → AVOID.

WATCH mandate — null position fields

When verdict is hold, all position-specific fields are null in both decision and mandate — entry_low, entry_high, stop, size_pct, entry_zone, stop_loss, stop_basis, size_hint_pct, size_basis. Only horizon and invalidations are populated.

Payment tier

The endpoint uses the upto scheme — agents sign for the 250k DELU ceiling, but the handler settles based on the caller's DELU balance on Base.

Payment token: DELU — 0x7b0ee9dcb5c1d4d7cd630c652959951936512ba3 on Base (18 decimals).

Check payment_tier.settled_delu in the response body for what was actually charged — not the x402 authorization ceiling.

Response schema summary

• decision — flat header: action, conviction, direction, entry_low, entry_high, stop, size_pct, read
• ca, chain, oracle_version
• verdict — strong_buy | buy | hold | avoid | drop
• score — 0–100 fused cognition score
• confidence — 0–1 data quality and signal agreement
• drivers / risks — up to 3 each
• signals — momentum, flow, structure, volatility, liquidity
• context — regime_label, regime_confidence, base_eco_pulse, macro_pulse
• mandate — action, entry_zone, stop_loss, stop_basis, size_hint_pct, size_basis, horizon, invalidations
• payment_tier — tier, delu_balance, settled_delu, note
• observed — always present: market, regime, social, deluagent (scout/auditor/quant mirror with weights_used)
• summary, selected_timeframe, candle_count, pool_source, timestamp

See references/response-schema.md for the full field-by-field schema. See references/mandate-fields.md for mandate construction details. See references/example-response.md for a full annotated response example. See references/social-enrichment.md for the opt-in two-step social flow. See references/external-clients.md for standalone client recipes.

Error codes