rugguard-sniper-bot-example

⚠ Educational, not production-grade. Read Safety before going live on mainnet.

An example Base sniper bot that demonstrates the canonical pre-trade safety integration with RugGuard. For each candidate token, the bot calls /v1/pretrade/check, then routes to a mock buy (sized by RugGuard's max_suggested_exposure_usd) or a skip log line based on policy_recommendation.

The bot does not actually execute on-chain. It logs what it would buy. Swap execute_buy_mock for your own DEX router call to make it real.

Why

Sniper bots that buy fresh token launches are textbook rug-pull bait. A pre-trade safety gate cuts a measurable chunk of losses before they happen. This kit shows the smallest-viable integration: ~120 LOC of bot logic on top of RugGuard's /v1/pretrade/check endpoint.

Install + offline demo

pip install rugguard-sniper-bot-example
rugguard-sniper --demo

The demo runs the bot through 3 representative candidates (safe USDC, fresh memecoin, drained pool) without paying or touching the network.

Expected output:

Sniper bot: 3 candidates on base, policy=balanced, intended_size=$100.0, session cap=$1.0

  [BUY  ] base:0x83358... score= 12 verdict=safe         size=$ 100.00/$100 reasons=[-] sig=a0c71156
  [BUY*] base:0x4ed4E... score= 62 verdict=medium_risk  size=$  20.00/$100 reasons=[OWNER_NOT_RENOUNCED, TOP10_CONCENTRATION_HIGH] sig=a0c71156
  [SKIP ] base:0xfc048... score= 95 verdict=critical    size=$   0.00/$100 reasons=[TOP10_CONCENTRATION_HIGH, LP_INSUFFICIENT_LIQUIDITY] sig=a0c71156

Session summary:
  candidates evaluated: 3
  executed (mock):      2
  skipped:              1
  RugGuard spend:       $0.03 USDC
  Would-be buy total:   $300.00
  Actual exposure:      $120.00
  Capital protected:    $180.00

The [BUY*] asterisk means "RugGuard clamped this from $100 to $20 due to medium risk". The Capital protected line shows the difference between naive-buy size and RugGuard-clamped size.

Run with real candidates (live x402 payment)

# 1. Get a funded x402 wallet (Base mainnet, ≥ $0.05 USDC)
export RUGGUARD_X402_PRIVATE_KEY=0xYOUR_PRIVATE_KEY_HEX

# 2. Inline list
rugguard-sniper --addresses 0xABC...,0xDEF... --chain base --size 100 --policy balanced

# 3. Or from a file (one address per line, '#' comments OK)
rugguard-sniper --addresses-file tokens.txt --chain base --size 100

Each candidate costs $0.01 USDC. The bot enforces a hard --session-cap (default $1) and aborts before exceeding it.

CLI flags

Policy modes (recap)

uncertain verdict → caution in all modes.

Library API

bot.py exports a few primitives you can call directly without using the CLI:

from rugguard_sniper_bot import (
    evaluate_candidate, run_sniper, SniperDecision, SniperStats, DecisionCache,
)

decision = await evaluate_candidate(
    chain="base",
    contract="0xABC...",
    intended_trade_usd=100.0,
    policy="balanced",
    private_key_hex=YOUR_KEY,
    api_url="https://rugguard.redfleet.fr",
    cache=DecisionCache(),
)
# decision.recommendation in {"allow", "caution", "block", "error"}
# decision.executed_size_usd  ← clamped by RugGuard

Safety

This is the most important section. Read before going live on mainnet.

What this kit does NOT do (and what you must add)

• No actual on-chain execution. execute_buy_mock is a log line. Swap it for your Uniswap V2 / Aerodrome router integration before going live.
• No mempool subscription. A real sniper listens for PoolCreated / PairCreated events. Add a websockets-based listener (Alchemy / QuickNode / Helius).
• No position management. No stop-loss, take-profit, time-based exit. Build a separate position_manager module.
• No retry / circuit breaker. Network errors route to recommendation="error" and skip. Production should retry transient failures with backoff, alert on sustained failures.
• No alerts / structured logging. print is the only output. Wire structlog + a Telegram/Discord webhook before production.

Recommended hardening

• Use rugguard-mcp for production x402. The bundled x402_pay.py here is pedagogical — it has an asset whitelist but no spend caps. The full rugguard-mcp ships session caps, 24h caps, file-locked TOCTOU-safe charge reservation, and replay-window protection.
• Dedicated wallet. Generate a fresh EOA, fund only with the USDC you can afford to lose. Treat the wallet balance itself as a hard cap.
• Verify signed reports. Production: install rugguard-verify and validate every SniperDecision's signature_fingerprint against /v1/pubkey before trusting it. The fingerprint is already surfaced in SniperDecision.signature_fingerprint.
• Independent sanity check. Don't trust RugGuard alone. Cross-check with at least one other signal (your own liquidity threshold, a different rug-scanner) for the trades that actually move money.

Wallet hygiene

# DO NOT commit the private key
echo "RUGGUARD_X402_PRIVATE_KEY=0x..." >> .env
echo ".env" >> .gitignore

The bot reads the key from the env var only. Never accept it as a CLI flag (would leak into shell history + ps output).

Companion packages

• rugguard-mcp — MCP server for Claude Desktop / Cursor / LangGraph
• rugguard-pydantic-ai-agent — Pydantic AI typed tool
• rugguard-langgraph-agent — LangGraph node
• rugguard-verify — Ed25519 signed-report verifier CLI

License

MIT. See LICENSE. Use at your own risk — this is example code, not investment advice.