Skip to content

galleonlabs/crypto-defi-skills / packages/lending/skills/galleon-aave-v4

galleon-aave-v4

Verified

Use when inspecting an Aave V4 Hub/Spoke position, sizing supply or borrowing, preparing repay/withdraw calls, or reconciling a V4 execution plan through official Aave MCP or AaveKit. V3 Pools, eMode and Aptos require their own interfaces.

FileKindSize
ATTRIBUTION.mdmarkdown946 B
LICENSEother1.0 KB
SKILL.mdskill md4.9 KB
agents/openai.yamlconfig236 B
references/examples.mdmarkdown1.9 KB
references/operations.mdmarkdown3.3 KB

references/operations.md

View on GitHub
# V4 official interface procedure

Primary documentation reviewed 2026-10-08. Installed tool schemas and observed chain state govern actual use.

## Tool discovery and selectors

[Aave MCP getting started](https://aave.com/docs/mcp/getting-started) documents the keyless HTTP endpoint `https://mcp.aave.com`, root POST and `/mcp` alias. The server supplies `tools/list`; use it as the argument authority. It does not offer an SSE GET stream. Clients using the 2026 protocol revision need its documented `Mcp-Method`/`Mcp-Name` headers; use the existing client's negotiated transport rather than inventing a handshake. Shared limits can produce 401/429.

[Aave tools](https://aave.com/docs/mcp/tools) expose V4 `get_hubs`, `get_hub_assets`, position items and operations-aware confirmation. Supply/borrow reserve selectors come from current `get_markets`; withdraw/repay selectors come from `get_position_items` for the exact owned position. Opaque reserve IDs differ from Spoke, Hub-asset and user-position IDs; preserve each in its own argument namespace. Parse the successful envelope from `data` and inspect warnings. Main-unit decimal amounts differ from raw contract inputs. Supply/APY context can share Hub accounting across spokes; do not call those independent sources.

## AaveKit implementation route

[V4 TypeScript](https://aave.com/docs/aave-v4/getting-started/typescript) still documents `@aave/client@next`, but on 2026-11-15 that tag was still a March 2026 prerelease (`4.0.0-next.54`) while `latest` was the V4 client 7.1.0 from `aave/aave-v4-sdk`. V3 uses the 0.x line from `aave/aave-sdk`. Pin a reviewed compatible version when implementing. Read-only client actions need no wallet. Await `ResultAsync`, branch on `isOk()`/`isErr()`, and preserve the actual failure instead of converting it to empty positions. Display overrides change names/icons, not the underlying token identity.

[Borrow](https://aave.com/docs/aave-v4/positions/borrow), [repay](https://aave.com/docs/aave-v4/positions/repay) and [withdraw](https://aave.com/docs/aave-v4/positions/withdraw) carry their own conditions and plans. Inspect the SDK's current union rather than assuming every result is a `TransactionRequest`. A balance failure is not approval to fund the wallet.

## Plan branches

The MCP action union documents `TransactionRequest`, `Erc20ApprovalRequired`, `PreContractActionRequired` and `InsufficientBalanceError`. An approval may offer `bySignature`; if the authorized user signs it, rebuild with `permitSignature` and the message's actual `permitDeadline`. Otherwise wait for the `byTransaction` receipt and visible allowance. Never fabricate a permit argument on the first call.

Native gateway plans have `transaction` before `originalTransaction`. Preserve order and check each payload. V4's `get_transaction_processed` consumes the exact `operations` returned by preparation. Corroborate a receipt independently; the activity feed cannot supply arbitrary receipt, gas or block information.

## Liquidation scope

[V4 liquidations](https://aave.com/docs/aave-v4/positions/liquidations) use health-dependent bonus and residual-debt constraints. Do not import a V3 fixed close-factor calculation. This skill's default is position planning; a liquidation request additionally needs the actual eligible debt/collateral, receiving account, simulation and separately authorized spending.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 323kUpdated Claude CodeCodexVerifiedInstall