# Segment 2 prep — Spellbook PR draft (Limitless Exchange, Base)

**Produced during Segment 1 Phase 3.** Candidate C2 deliverable — a
Dune Spellbook PR that's ready to submit the moment `merovan` has a working
GitHub account.

## Headline

Drafted a complete Spellbook spell set for **Limitless Exchange** (largest
prediction-market platform on Base, ~$1.3B cumulative volume, CLOB-based).
Limitless is completely unspelled in `duneanalytics/spellbook` as of
2026-04-18 — verified by `find spellbook/ -iname '*limitless*'` (no
matches). This is a legitimate high-signal first PR opportunity for the
`merovan` handle.

## Files drafted

Location: `scripts_segment_2_prep/spellbook_draft/limitless_base_spell/`

```
sources/limitless/base/_sources.yml          -- source table declarations
models/base/_schema.yml                       -- model metadata + contributors
models/base/limitless_base_market_trades_raw.sql  -- raw OrderFilled events
models/base/limitless_base_market_trades.sql      -- analyst-facing spell
```

Plus `scripts_segment_2_prep/spellbook_draft/README.md` with complete
submission instructions.

## Key design choices

- **Protocol choice: Limitless over Polymarket or Kalshi.** Polymarket +
  Kalshi both already have extensive Spellbook coverage (19 spells on
  Polymarket alone). Limitless has zero. Best marginal contribution.
- **Pattern: CLOB OrderFilled unions.** Limitless uses the same Gnosis CTF
  + versioned CTFExchange architecture as Polymarket, so the
  `polymarket_polygon_market_trades_raw` spell pattern transferred
  directly. Six source-event tables (simple v1/v2/v3 + NegRisk v1/v2/v3)
  are unioned into one raw-trades table.
- **Scope: trades first, details later.** A single well-tested `trades`
  spell is a better first PR than a sprawling patch. Follow-up patches
  add `market_details`, `ohlcv_hourly`, `positions`.
- **Contributors metadata set to `merovan`.** Reputation attached to every
  query that references `limitless_base.market_trades`.

## Headline findings

- Limitless Exchange contract addresses (all on Base, verified from
  `docs.limitless.exchange/user-guide/smart-contracts` on 2026-04-18):
  - CTF: `0xC9c98965297Bc527861c898329Ee280632B76e18`
  - Six CTFExchange variants (simple v1/v2/v3 + NegRisk v1/v2/v3)
  - Three NegRiskAdapter variants
- Dune needs each of these decoded first (via the submit-contract UI —
  independent of the Spellbook PR flow).
- Spell draft compiles as YAML (pyyaml `safe_load` passes on both yml
  files; SQL templated with dbt Jinja is syntactically plausible).

## Testing done

- YAML syntax: `pyyaml.safe_load` passes on `_sources.yml` and
  `_schema.yml`.
- SQL structure: mirrors `polymarket_polygon_market_trades_raw` (which
  is a merged, approved precedent in the spellbook repo).
- Cross-reference: contract-address list matches Limitless's canonical
  docs page.

### Added in Segment 2 Phase 0 (`segment_2_prep_and_poll`)

The Phase-3 recommendation flagged running `dbt compile` on this spell
as a must-do-first blocker before authoring any more spell drafts.
Completed 2026-04-18:

1. **`dbt parse` on the full daily_spellbook project** (with Limitless
   spells placed in `dbt_subprojects/daily_spellbook/models/_projects/
   limitless/base/` and `sources/limitless/base/_sources.yml`) — succeeded.
   Only warnings were deprecations inherited from unrelated repo models
   (`DuplicateYAMLKeysDeprecation`, `MissingArgumentsPropertyIn
   GenericTestDeprecation`, `PropertyMovedToConfigDeprecation`); no
   error from any Limitless file.
2. **`dbt compile --select limitless_base_market_trades_raw
   limitless_base_market_trades`** — succeeded. The Jinja templating
   (including the `{{ source('limitless_base', ...) }}` references, the
   `{{ is_incremental() }}` / `{{ incremental_predicate(...) }}`
   macros, and the `expose_spells` post-hook) rendered to clean SQL:
   `"delta_prod"."limitless_base"."CTFExchange_v1_evt_OrderFilled"`-style
   fully-qualified references.
3. **`sqlglot.parse(..., dialect='trino')`** on both compiled outputs
   (`target/compiled/daily_spellbook/models/_projects/limitless/
   base/*.sql`) — both parse successfully (1 statement each).
4. **No column-name surprises** — the `makerAssetId` vs `takerAssetID`
   case mismatch in the UNION columns is a known Solidity-source typo
   inherited from Polymarket's CTFExchange (verified present in the
   approved `polymarket_polygon_market_trades_raw` spell), not a bug in
   our spell. We carry it forward for consistency because Limitless
   forks Polymarket's CTFExchange 1:1.

**Caveats — things `dbt compile` + `sqlglot` cannot catch without a
Catalyst workspace:**
- Schema-level column existence: dbt doesn't verify that
  `limitless_base.CTFExchange_v1_evt_OrderFilled` actually has a
  `makerAssetId` column once Dune's decoder registers the contract.
  This depends on the decoder-submission step (still gated on Dune
  account).
- `dbt-utils.unique_combination_of_columns` test execution — runs
  against live data only.
- `expose_spells` post-hook side effects — only run at materialization.

**Recommendation**: this spell is dbt-compile-clean and ready to submit
when the GitHub account exists. Do NOT publish the PR before the
decoder-submission flow places the six CTFExchange contracts into
Dune's `limitless_base` namespace — the spell is correct but will fail
CI if sources don't exist yet.

### Skipped this phase (per Step 0.7 classification (c))

INSTRUCTIONS.md § B2 directs: if the Dune-anon-submission assumption
classifies as (c) "compound-prep thesis broken," drop B2 to a
minimum-viable placeholder (finish Limitless validation, skip the new
Spellbook PR draft, redirect time to B3 brainstorm + bypass-path
evaluation).

Step 0.7 classified as (c) — see
`writeups/dune_submission_assumption_check.md`. Therefore the second
Spellbook PR draft planned by default in B2 is deferred; the
dbt-compile validation above closes the must-do-first gap without
stacking more speculative artifacts in a contracting review queue.

## Not done (explicitly deferred)

- **Actually submitting the PR** — requires GitHub account, which doesn't
  exist yet in this phase.
- **Follow-up spells** (`market_details`, `ohlcv_hourly`, `positions`,
  `users`) — structure known; implementation deferred until first PR
  lands successfully.
- **Second Spellbook PR draft** — deferred per Step 0.7 (c)
  classification, see above.

## Value of this artifact

1. **Head-start when GitHub exists.** PR submission is ~1 hour of work
   (fork + copy + pre-commit + push + PR body) instead of 4-6h of
   authoring from scratch.
2. **Independent research value.** The contract-address table itself is
   reusable by any Limitless analytics dashboard Segment 2 builds.
3. **Backup if GitHub signup stays stuck.** The draft can also be proposed
   informally to the Dune team via the Discord bounty channel with a
   pointer to this repo once we have Discord.
4. **Reputation compound.** If merged, every `limitless_base.*` query
   forever references `contributors: merovan`.

## Clone management

`duneanalytics/spellbook` shallow-cloned to
`scripts_segment_2_prep/spellbook_draft/spellbook/` (263MB). Added to
`.gitignore`. Used for pattern-matching only; not modified.