Yields Llama

Query DeFi yield opportunities across chains using the yieldsllama CLI (powered by DeFi Llama API). Use when the user asks about DeFi yields, APY, best staki...

Security Scan

The skill's instructions, requirements, and behavior are consistent with a CLI that queries DeFi Llama; it does require building and installing third‑party code and will write files in the working directory, so review before running.

Purpose & Capability

The name/description (querying DeFi yields via DeFi Llama) matches the SKILL.md: it documents a CLI that queries yields.llama.fi, filters and sorts results, and caches responses. Requiring a built binary and an LLAMA_DOMAIN endpoint is coherent for this purpose.

Instruction Scope

Instructions include building from source, creating a .env file in the current working directory (LLAMA_DOMAIN), and caching API responses to data.json. These actions are within the CLI's scope but will create files in whatever directory the agent runs in (possible surprise/side effects). The SKILL.md does not instruct reading unrelated user files or secrets.

Install Mechanism

There is no formal install spec in the registry, but SKILL.md instructs running the official rustup installer (https://sh.rustup.rs) and git-cloning a GitHub repo (https://github.com/0x9bb1/yieldsllama.git), then building and copying a binary into /usr/local/bin. The hosts used are known (rustup.rs and github.com) which reduces risk, but these steps implicitly run remote code and require elevated permissions to install into system PATH.

Credentials

Registry metadata lists no required env vars, but the runtime instructions require a .env file with LLAMA_DOMAIN (not a secret) and the tool will contact yields.llama.fi. No credentials or unrelated secrets are requested. The metadata/instructions mismatch (no declared env but .env required at runtime) is a minor inconsistency to be aware of.

Persistence & Privilege

always is false and the skill does not request permanent agent privileges. However installation instructions suggest placing a binary in /usr/local/bin (system-wide) and the tool creates a local cache file data.json; both are persistent changes to the host filesystem and may require root to perform.

Assessment

This skill appears to do what it claims, but it requires building and installing third‑party code and will create files in the directory where it runs. Before installing or running: (1) review the GitHub repository source (https://github.com/0x9bb1/yieldsllama) to ensure you trust it; (2) avoid running curl | sh blindly — prefer installing rustup from official docs or using a package manager you control; (3) run the build/install steps inside a container or VM if you want to limit system impact; (4) be aware the install copies a binary into /usr/local/bin (may need sudo) and the tool will create/modify a .env and data.json in the current working directory — run it from a safe/empty directory to avoid overwriting project config; (5) confirm network calls to yields.llama.fi are acceptable in your environment. If you want stronger assurance, ask the publisher for a release tarball + checksum or a prebuilt binary from a vetted source.

Query and compare DeFi yield pool data from DeFi Llama.

Installation

Prerequisites

• Rust toolchain (cargo): https://www.rust-lang.org/tools/install

If Rust is not installed, run:

curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
source "$HOME/.cargo/env"

Build and Install

# Clone the repo
git clone https://github.com/0x9bb1/yieldsllama.git /tmp/yieldsllama

# Build release binary
cd /tmp/yieldsllama && cargo build --release

# Install to PATH
cp /tmp/yieldsllama/target/release/yieldsllama /usr/local/bin/yieldsllama

# Verify
yieldsllama --help

Environment Setup

The tool needs a .env file in the current working directory with the DeFi Llama API endpoint. Before each run, ensure it exists:

if [ ! -f .env ]; then echo 'LLAMA_DOMAIN="https://yields.llama.fi"' > .env; fi

CLI Usage

yieldsllama [OPTIONS]

Options

Flag	Description	Default
-l, --limit <N>	Number of results to return	10
-e, --exposure <TYPE>	Pool type: single or multi	single
-s, --sort <FIELD>	Sort by: apy or tvl	apy
-t, --tvl <MIN>	Minimum TVL in USD	none
-c, --chain <CHAIN>	Filter by chain name (case-insensitive)	none
-a, --asset <TOKEN>	Filter by token symbol (case-insensitive)	none

Common Examples

# Top 10 highest APY single-asset pools
yieldsllama

# Top 20 pools sorted by TVL
yieldsllama -l 20 -s tvl

# Best USDC yields with TVL > $10M
yieldsllama -a USDC -t 10000000

# Best USDT yields on Ethereum
yieldsllama -a USDT -c Ethereum

# Best multi-asset pool yields
yieldsllama -e multi -l 10

# Compare stablecoins with TVL > $10M
yieldsllama -a USDC -t 10000000 -l 5
yieldsllama -a USDT -t 10000000 -l 5
yieldsllama -a USDE -t 10000000 -l 5
yieldsllama -a DAI -t 10000000 -l 5

# Best ETH yields with TVL > $1M
yieldsllama -a WETH -t 1000000 -l 10

# Best yields on specific chains
yieldsllama -c Ethereum -t 1000000 -l 10
yieldsllama -c Base -t 1000000 -l 10
yieldsllama -c Arbitrum -t 1000000 -l 10

Output Columns

Column	Description
apy	Total APY (apyBase + apyReward)
symbol	Token symbol
chain	Blockchain network
project	Protocol name
tvlUsd	Total Value Locked in USD
apyBase	Base lending/staking APY (organic, sustainable)
apyReward	Reward token APY (may be unsustainable, subject to token price)
exposure	single (one asset) or multi (LP pair etc.)

Guidance for Analyzing Results

When presenting yield data to the user, always consider and mention:

1. apyBase vs apyReward: Base APY is from lending/staking fees (more sustainable). Reward APY comes from incentive tokens (can drop to 0 when rewards end or token dumps).
2. TVL matters: Very high APY with low TVL (<$1M) is often unsustainable or risky. Recommend filtering with -t for serious allocation decisions.
3. APY > 100% is suspicious: Flag any extremely high APY and warn the user about sustainability.
4. Chain risk: Different chains have different security profiles. Ethereum L1 is generally safest; newer L2s/alt-L1s carry more risk.
5. Protocol risk: Well-known protocols (Aave, Compound, Morpho) are generally safer than unknown ones.

Workflow for Multi-Token Comparison

When the user asks "which token/pool has the best yield", run queries for each token in parallel and present a unified comparison table. Example:

User: "Compare USDC, USDT, USDE yields with TVL > $10M"

Run in parallel:
  yieldsllama -a USDC -t 10000000 -l 5
  yieldsllama -a USDT -t 10000000 -l 5
  yieldsllama -a USDE -t 10000000 -l 5

Then summarize the top option for each token and give a recommendation.

Refreshing Data

The tool caches API responses in data.json for 12 hours. To force refresh:

rm -f data.json && yieldsllama

Troubleshooting

Problem	Solution
yieldsllama: command not found	Binary not in PATH. Run cp /path/to/yieldsllama /usr/local/bin/
读取本地data.json文件异常	Normal on first run — it fetches from API and creates data.json
本地文件过期	Cache expired (>12h). Tool auto-fetches fresh data
Empty results	Token symbol or chain name may not match exactly. Try without -c/-a filters first
Network error	Check internet connection; DeFi Llama API (yields.llama.fi) may be temporarily down
.env error	Ensure .env exists in current directory with LLAMA_DOMAIN="https://yields.llama.fi"