agentleFS
agiprolabsagiprolabs/claude-trading-skills/skills/shredstream/SKILL.md

shredstream

Skill366 starsChanged 8 months agoMITView on GitHub

Pre-execution Solana transaction streaming via Jito ShredStream, Shyft RabbitStream, and Triton Deshred

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Wait for approval (your keypair gets whitelisted)

    line 59
  2. Distinctive:

    No staking requirement, free during beta

    line 60
  3. Very specific:

    Limitations: Only transaction filters work.

    line 178

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

ShredStream — Pre-Execution Solana Data

ShredStream gives you transaction data before the validator executes the block — typically 100-500ms earlier than standard Yellowstone gRPC. You see transaction intent, not confirmed results.

This is the fastest path to Solana data for time-critical trading strategies.

How It Works

Solana validators produce blocks by serializing transactions into shreds (~1,228 bytes each, sized for UDP MTU). Shreds propagate through Turbine (Solana's fanout protocol, 2-3 hops). ShredStream bypasses Turbine by receiving shreds directly from leader validators via Jito's Block Engine.

Leader Validator
     │
     ├── Turbine (standard, 2-3 hops, 200-500ms)
     │       └── Your RPC Node → Yellowstone gRPC (post-execution)
     │
     └── Jito Block Engine (direct)
              └── ShredStream Proxy (your server)
                   ├── UDP shreds → Your RPC/Validator (faster block building)
                   └── gRPC entries → Your Trading Bot (decoded transactions)

What You Get vs. What You Don't

Available (Pre-Execution)NOT Available (Needs Execution)
Transaction signaturesSuccess/failure status
Account keys (pubkeys)Balance changes (pre/post)
Instructions (program, accounts, data)Log messages
Address lookup table referencesInner instructions (CPI)
Slot numberToken balance changes
Compute units consumed

Key tradeoff: Speed for completeness. You see what's about to happen but can't confirm it actually succeeded. Some transactions you see will ultimately fail.

Three Ways to Get Pre-Execution Data

ProviderProductLatencyAccessCost
JitoShredStream Proxy~10-50ms from leaderApply + auth keypairFree (beta)
ShyftRabbitStream~15-100ms faster than gRPCShyft gRPC planFrom $199/mo
TritonDeshred (SubscribeDeshred)~6.3ms p50 from shredTriton customer~$2,900+/mo

See references/providers_compared.md for detailed comparison.

Option 1: Jito ShredStream Proxy

The most direct approach — run Jito's open-source proxy on your own server.

Get Access

  1. Generate a Solana keypair: solana-keygen new -o shred_auth.json
  2. Apply at Jito's form with your public key
  3. Wait for approval (your keypair gets whitelisted)
  4. No staking requirement, free during beta

Run the Proxy

# Clone and build
git clone https://github.com/jito-labs/shredstream-proxy.git --recurse-submodules
cd shredstream-proxy

# Run with gRPC enabled (key flag: --grpc-service-port)
RUST_LOG=info cargo run --release --bin jito-shredstream-proxy -- shredstream \
    --block-engine-url https://mainnet.block-engine.jito.wtf \
    --auth-keypair /path/to/shred_auth.json \
    --desired-regions ny,amsterdam \
    --dest-ip-ports 127.0.0.1:8001 \
    --grpc-service-port 7777

Docker (host networking required for UDP):

docker run -d --name shredstream-proxy --rm \
  --network host \
  -e RUST_LOG=info \
  -e BLOCK_ENGINE_URL=https://mainnet.block-engine.jito.wtf \
  -e AUTH_KEYPAIR=/app/shred_auth.json \
  -e DESIRED_REGIONS=ny,amsterdam \
  -e DEST_IP_PORTS=127.0.0.1:8001 \
  -e GRPC_SERVICE_PORT=7777 \
  -v /path/to/shred_auth.json:/app/shred_auth.json \
  jitolabs/jito-shredstream-proxy shredstream

Configuration

ParameterDescriptionExample
BLOCK_ENGINE_URLJito block engine endpointhttps://mainnet.block-engine.jito.wtf
AUTH_KEYPAIRPath to whitelisted Solana keypairshred_auth.json
DESIRED_REGIONSMax 2, comma-separatedny,amsterdam
DEST_IP_PORTSWhere to forward raw shreds (UDP)127.0.0.1:8001
GRPC_SERVICE_PORTEnable gRPC entry streaming7777
SRC_BIND_PORTIncoming shred UDP port20000

Available regions: amsterdam, dublin, frankfurt, london, ny, salt-lake-city, singapore, tokyo

Verify It's Working

# Check shreds are arriving via UDP
sudo tcpdump 'udp and dst port 20000'
# Should see many ~1200-byte packets continuously

Consume via gRPC

use jito_protos::shredstream::{
    shredstream_proxy_client::ShredstreamProxyClient,
    SubscribeEntriesRequest,
};

let mut client = ShredstreamProxyClient::connect("http://127.0.0.1:7777").await?;
let mut stream = client
    .subscribe_entries(SubscribeEntriesRequest {})
    .await?
    .into_inner();

while let Some(entry) = stream.message().await? {
    let entries: Vec<solana_entry::entry::Entry> =
        bincode::deserialize(&entry.entries)?;

    for e in &entries {
        for tx in &e.transactions {
            let sig = tx.signatures[0];
            let msg = tx.message();
            // Parse instructions, accounts, etc.
        }
    }

    println!("Slot {}: {} entries, {} transactions",
        entry.slot,
        entries.len(),
        entries.iter().map(|e| e.transactions.len()).sum::<usize>()
    );
}

Option 2: Shyft RabbitStream

Drop-in replacement for Yellowstone gRPC — same SubscribeRequest format, just a different endpoint. Easiest way to get pre-execution data without running infrastructure.

export GRPC_ENDPOINT="https://rabbitstream.ny.shyft.to"
export GRPC_TOKEN="your-shyft-x-token"
# Same code as yellowstone-grpc, just different endpoint
import grpc

endpoint = "rabbitstream.ny.shyft.to"
token = os.environ["GRPC_TOKEN"]

# ... standard Yellowstone connection code ...
# Subscribe to transactions — same filter format
request = SubscribeRequest(
    transactions={
        "pumpfun": SubscribeRequestFilterTransactions(
            account_include=["6EF8rrecthR5Dkzon8Nwu78hRvfCKubJ14M5uBEwF6P"],
            vote=False,
            failed=False,
        )
    },
    commitment=CommitmentLevel.PROCESSED,
)

Limitations: Only transaction filters work. No account, slot, or block subscriptions. The meta field is empty (no execution results).

Regional endpoints: rabbitstream.{ny,va,ams,fra}.shyft.to

Option 3: Triton Deshred

Lowest latency (~6.3ms p50) via Triton's SubscribeDeshred RPC. Same Yellowstone client, different method.

// Requires yellowstone-grpc-client with Deshred support
let (mut tx, mut stream) = client.subscribe_deshred().await?;

tx.send(SubscribeDeshredRequest {
    deshred_transactions: hashmap!{
        "pumpfun".to_string() => SubscribeRequestFilterDeshredTransactions {
            vote: Some(false),
            account_include: vec!["6EF8rrecthR5Dkzon8Nwu78hRvfCKubJ14M5uBEwF6P".into()],
            ..Default::default()
        }
    },
    ..Default::default()
}).await?;

Access: Triton customers only, paid beta, requires their custom Agave validator fork.

Parsing Pre-Execution Transactions

Without execution metadata, parsing is simpler but requires program-specific knowledge.

Identify the Program

# From a raw VersionedTransaction (post-deserialization)
account_keys = [str(k) for k in tx.message.account_keys]
for ix in tx.message.instructions:
    program_id = account_keys[ix.program_id_index]
    if program_id == "6EF8rrecthR5Dkzon8Nwu78hRvfCKubJ14M5uBEwF6P":
        # This is a PumpFun instruction
        discriminator = ix.data[:8]
        # Decode instruction data per PumpFun IDL

Instruction Discriminators

Most Solana programs use 8-byte discriminators (Anchor SHA256 hash of the instruction name). Match instruction.data[:8] against known values for each program.

# Common approach
PUMPFUN_CREATE = bytes.fromhex("181ec828051c0777")
PUMPFUN_BUY = bytes.fromhex("66063d1201daebea")
PUMPFUN_SELL = bytes.fromhex("33e685a4017f83ad")

disc = ix.data[:8]
if disc == PUMPFUN_BUY:
    # Parse buy parameters from remaining bytes
    ...

Warning: Discriminator values are program-specific and can change between program versions. Always verify against the current program IDL. See the pumpfun-mechanics skill for PumpFun-specific parsing.

Common Architecture: ShredStream + Yellowstone

Most production systems use both:

ShredStream (pre-execution)          Yellowstone gRPC (post-execution)
         │                                      │
         ▼                                      ▼
  Intent Detection                    Confirmation + Reconciliation
  "Wallet X is buying token Y"       "Buy succeeded, wallet now holds Z"
         │                                      │
         ▼                                      ▼
  Pre-compute Response                Execute / Update State
  (route, sign, prepare bundle)       (record PnL, update positions)

This gives you the speed advantage of ShredStream for signal detection while using Yellowstone for reliable state management.

Deployment Requirements

  • Public IP required — NAT breaks UDP shred delivery
  • Host networking — Docker bridge mode drops shred packets
  • UDP port 20000 open for incoming shreds
  • Co-locate near validators — Frankfurt, NY, Amsterdam, London, Tokyo recommended
  • Bare metal preferred — Cloud VMs add 1-5ms jitter from shared NICs

See references/deployment.md for full infrastructure guide.

Files

References

  • references/providers_compared.md — Jito ShredStream vs Shyft RabbitStream vs Triton Deshred
  • references/deployment.md — Infrastructure requirements, region selection, firewall configuration
  • references/proto_reference.md — ShredStream protobuf definitions and Entry parsing

Scripts

  • scripts/parse_shredstream_entries.py — Decode and analyze ShredStream gRPC entries
  • scripts/rabbitstream_monitor.py — Connect to Shyft RabbitStream for pre-execution transaction monitoring

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.