One-prompt Substreams generation attempt (Task 11, Step 2)

Status: skipped. Per the controller's resolution for Task 11, installing the Substreams Skills plugin (claude plugin marketplace add streamingfast/substreams-skills && claude plugin install substreams-dev@streamingfast-substreams) was out of scope for this run: it requires installing a Claude Code plugin, which the task explicitly said not to do. This step is optional and cheap to redo later, so it was skipped rather than blocking the rest of Task 11 on it.

The scaffold in substreams/erc4626-vault-metrics/ was instead built by hand, following Steps 3-7 of the task brief directly.

The prompt, for whoever runs this later

If the Substreams Skills plugin is installed later, this is the exact one-prompt text specified by the task brief. Run it in a fresh Claude Code session inside substreams/, save the output under version control as chore(substreams): one-prompt generation (unreviewed), and record the screen:

Create a Substreams package erc4626-vault-metrics for Ethereum mainnet that imports the Pinax erc4626 package from https://github.com/pinax-network/substreams-evm/raw/main/spkg/erc4626-v0.1.0.spkg, maps its Deposit and Withdraw events to a VaultEvent proto with implied share price = assets/shares, and emits a SQL sink db_out with a vault_events table.

What the hand-built scaffold covers instead

substreams/erc4626-vault-metrics/ already has a working manifest, protobuf schema, and map_vault_events module (see the Task 11 report for full detail: .superpowers/sdd/2026-09-09-vaultradar/task-11-report.md). Note one concrete finding relevant to the prompt above: the sql sink half of the one-prompt ask (db_out / SQL sink) hit a real blocker independent of any prompt wording. The only available substreams-sink-sql protodefs release, protodefs-v1.0.7 (Nov 2023), fails substreams protogen outright under the current substreams CLI (1.22.0) with a self-contained proto extension conflict, before any package code is even involved. A one-prompt attempt asking for a db_out sink would hit the same wall. Task 13 should build DatabaseChanges with the substreams-database-change crate (crates.io, v4.x) instead of importing that spkg.