Data And Intelligence

Data posture

Raw provider payloads are not reliable enough to become agent truth directly. Providers disagree, fields are missing, freshness varies, and textual summaries can hide important gaps.

The repo direction is to normalize external data into typed contracts before handing it to the model layer.

Key contracts

• CanonicalAnalysisSnapshot
• DecisionFeatureBundle
• MarketContextPack
• provider-specific fundamental or news snapshots before canonical merge

Why this matters

The model should not be asked to infer unstated finance truth from vague strings. It should receive:

• structured summaries
• missing-field visibility
• freshness context
• source attribution
• explicit risk flags

Provider aggregation rules

Good aggregation in this repo means:

• prefer canonical output over provider-specific string parsing
• keep missing provider evidence visible
• preserve source attributions
• degrade confidence when the evidence layer is thin
• keep generic fallbacks clearly labeled as fallbacks

Feature preparation

The feature layer is where the repo turns mixed market, provider, and runtime signals into a compact decision context. If this layer is missing or incomplete, downstream confidence should reflect that.

That is why missing context or missing decision_features should be treated as a real provider or evidence gap rather than a cosmetic omission.

News and world-state evidence

News, filings, macro releases, browser fetches, and search results are useful only after they become small evidence packets. A usable packet should say where it came from, when it was published, when it was fetched, whether it was fresh or stale, what fetcher path was used, what source tier it belongs to, and why it is material.

Agentic Trader should not hand raw article text or social posts to trading prompts. The safer path is: collect, redact, summarize, classify, cite, then let scanner and proposal review surfaces consume the compact result.

The first operator-facing runtime surface for this is:

agentic-trader news-intelligence --symbol AAPL --company-name Apple --sector "consumer technology" --json

It does not scrape the web by itself. It produces query templates, source tiers, freshness rules, excluded noisy domains, material event types, and the prompt policy that keeps raw article text out of the core trading prompt.

Strategy intelligence

The V1 strategy path is intentionally conservative. Momentum, gap review, mean-reversion, breakout/reclaim, and volatility presets are research helpers until they are enriched with provider evidence, liquidity/spread checks, sizing assumptions, and proposal review.

Those profiles are now available as runtime metadata:

agentic-trader strategy-catalog --json
agentic-trader strategy-profile vwap-breakout --json
agentic-trader idea-score --symbol AAPL --preset momentum --price 190 --volume 5000000 --change-pct 6.2 --relative-volume 3.4 --json
agentic-trader proposal-candidate-create --symbol AAPL --preset momentum --price 190 --volume 5000000 --change-pct 6.2 --relative-volume 3.4 --quantity 1 --stop-loss 182 --take-profit 205 --json
agentic-trader proposal-candidate-promote CANDIDATE_ID --json

The output explains required inputs, missing evidence, risk controls, validation checks, and why scanner output still needs broker-free candidate review before it can become a pending proposal.

More advanced patterns such as opening-range, VWAP, Keltner/Bollinger, regime-adaptive, pairs, or ensemble ideas should first pass no-lookahead checks, backtest confidence review, and clear invalidation rules. Even then, they create proposal candidates or pending proposals after promotion, not orders.

Finance/accounting evidence

finance-ops --json now carries a reconciliation contract for ledger categories such as trades, cash, fees/taxes, dividends, interest, and corporate actions. V1 paper mode can honestly mark some of those categories as missing; the important rule is that missing evidence must never be treated as zero evidence.

Continuous research cycle

The benchmark loop pattern is available as a read-only runtime contract:

agentic-trader research-cycle-plan --symbols AAPL,MSFT --json

The phases are PRE-FLIGHT, MONITOR, ANALYZE, PROPOSE, and DIGEST. The plan names the existing commands each phase reads and repeats the hard safety line: the cycle may surface ideas and pending proposals, but it may not approve, submit, mutate policy, or hide missing source evidence.

Review and persistence tie-in

These structured artifacts are not only for live reasoning. They also matter for:

• review surfaces
• QA
• replay
• debugging disagreement between providers and agents

Safe mental model

Think of the model as a reasoning layer sitting on top of typed evidence, not as the place where missing market truth gets invented.