Contract Auditor Skill — Buzz BD Agent v1.0

Fast security feedback on EVM token contracts. Findings in minutes, not weeks. Runs BEFORE /score-token scoring. Enriches safety sub-agent output with on-chain proof.

1. MISSION

When a token prospect arrives with a contract address, Buzz must NOT rely solely on RugCheck heuristics. This skill fetches the actual contract source and runs a structured security pass to surface high-confidence vulnerabilities that automated scanners miss or under-weight.

Time budget: < 3 minutes per contract
Output: Structured audit summary → piped into /score-token safety score
Revenue impact: Justifies premium /score-token pricing ($0.50–$1.00 via x402)

2. WHEN TO RUN

Run contract-auditor automatically when ANY of these conditions are met:

Trigger	Condition
`/score-token` called	Contract address provided + source is verified
Listing inquiry received	Prospect submits token for SolCex listing review
Safety sub-agent invoked	Scanner sub-agent passes contract to safety sub-agent
Manual request	Ogie asks Buzz to audit a specific contract address

Skip if:

Contract source is not verified on Etherscan/Sourcify (flag this as a RED signal)
Token is Solana SPL (use Helius + RugCheck instead — see Section 6)
Contract is > 5,000 lines (flag for manual audit referral)

3. STEP-BY-STEP WORKFLOW

Step 1 — Fetch Contract Source

GET https://api.etherscan.io/api
  ?module=contract
  &action=getsourcecode
  &address={CONTRACT_ADDRESS}
  &apikey={ETHERSCAN_API_KEY}

For Base chain:

GET https://api.basescan.org/api
  ?module=contract
  &action=getsourcecode
  &address={CONTRACT_ADDRESS}
  &apikey={BASESCAN_API_KEY}

If not verified: Stop. Flag CONTRACT_NOT_VERIFIED as HIGH risk signal. Add to score: -25 safety points. Report to Telegram and pipeline.

Step 2 — Count Lines / Triage

lines = source.split('\n').length
if lines > 5000: flag LARGE_CODEBASE, recommend manual audit, continue with partial scan
if lines < 50: flag SUSPICIOUSLY_SMALL, add to findings

Step 3 — Pattern Analysis Pass

Run ALL checks in Section 4. Log each finding with:

severity: CRITICAL | HIGH | MEDIUM | LOW | INFO
pattern: the vulnerability class
evidence: the specific function/line pattern found
confidence: HIGH | MEDIUM | LOW

Step 4 — Score Calculation

base_safety_score = 100
CRITICAL finding: -30 each (cap at -60)
HIGH finding:     -15 each (cap at -30)
MEDIUM finding:   -5 each
LOW finding:      -2 each
CONTRACT_NOT_VERIFIED: -25 flat
LARGE_CODEBASE:   -5 flat (uncertainty penalty)

Step 5 — Generate Report

Output the structured report (Section 5 format). Pipe contract_safety_score and findings[] into the /score-token response. Send Telegram summary to Ogie (@Ogie2) for any CRITICAL findings.

4. SECURITY CHECKS (Pattern Library)

Read /contract-auditor/references/patterns.md for the full pattern library.

Quick reference — CRITICAL checks (always run these first):

Check	Pattern to Find	Flag If...
Hidden Mint	`mint()`, `_mint()` outside constructor	Callable post-deploy by owner
Backdoor Transfer	`transferFrom` override, `_transfer` with owner bypass	Owner can move any wallet's tokens
Fee Trap	`_taxFee`, `_liquidityFee` > 25% or dynamically settable	Fees can be set to 100%
Blacklist	`blacklist[]`, `isBlacklisted`, `bots[]` mapping	Owner can freeze arbitrary wallets
Pausable	`pause()`, `whenNotPaused` modifier	Owner can halt all transfers
Upgrade Proxy	`upgradeTo()`, `_implementation`, `delegatecall`	Contract logic replaceable post-deploy
Ownership Not Renounced	`owner()` != address(0)	Single point of control remains
Rug via LP	`removeLiquidity` callable by owner without timelock	Instant LP drain possible

HIGH checks:

Check	Pattern	Flag If...
Max TX Manipulation	`_maxTxAmount` setter	Settable to 0 (trading halt)
Unchecked Return	Low-level `call()` without return check	Silent failure on ETH send
Reentrancy	State change AFTER external call	Classic reentrancy shape
Integer Overflow	Pre-0.8.0 Solidity without SafeMath	Any arithmetic on token amounts
tx.origin Auth	`require(tx.origin == owner)`	Phishing vulnerability
Centralized Price Oracle	Single `setPrice()` owner function	Price manipulation

5. OUTPUT FORMAT

{
  "contract_audit": {
    "address": "0x...",
    "chain": "ethereum|base|bsc",
    "verified": true,
    "source_lines": 847,
    "compiler": "v0.8.19",
    "scan_timestamp": "2026-03-06T12:00:00Z",
    "contract_safety_score": 72,
    "risk_level": "MEDIUM",
    "findings": [
      {
        "severity": "HIGH",
        "pattern": "BLACKLIST",
        "evidence": "function addToBlacklist(address account) external onlyOwner",
        "confidence": "HIGH",
        "description": "Owner can blacklist arbitrary wallet addresses, freezing their tokens"
      }
    ],
    "flags": ["OWNERSHIP_NOT_RENOUNCED"],
    "recommendation": "REVIEW_REQUIRED",
    "audit_summary": "Contract has standard ERC-20 structure with 1 HIGH finding (blacklist mechanism). Ownership not renounced. Recommend requesting renouncement before listing.",
    "listing_recommendation": "CONDITIONAL"
  }
}

listing_recommendation values:

APPROVE — No CRITICAL/HIGH findings, ownership renounced
CONDITIONAL — HIGH findings present but mitigable, or ownership not renounced
REJECT — CRITICAL findings, or contract not verified
ESCALATE — Unusual patterns requiring human review

6. SOLANA SPL TOKENS

Contract source auditing does NOT apply to Solana SPL tokens. For SPL tokens, use this alternative safety chain:

Helius — Check mint authority (mintAuthority: null = frozen = GOOD)
Helius — Check freeze authority (freezeAuthority: null = GOOD)
RugCheck (rugcheck.xyz/tokens/{mint}) — Full rug score
DexScreener — Liquidity lock status, LP burned %

Flag MINT_AUTHORITY_ACTIVE as HIGH risk if mintAuthority is not null. Flag FREEZE_AUTHORITY_ACTIVE as HIGH risk if freezeAuthority is not null.

7. INTEGRATION WITH BUZZ PIPELINE

REST API Endpoint

Audit results are stored in SQLite and exposed via:

GET /api/v1/audit/{address}
POST /api/v1/audit/run  { "address": "0x...", "chain": "base" }

Score-Token Integration

contract_safety_score from this skill feeds into /score-token response:

final_safety_score = (rugcheck_score * 0.4) + (contract_audit_score * 0.6)

Contract audit carries MORE weight than RugCheck because it's source-level.

JVR Receipt

After every audit, log to JVR receipt system:

AUDIT_COMPLETE | {address} | score:{N} | findings:{N} | {recommendation}

Send Telegram notification to Ogie for any CRITICAL findings or REJECT recommendations.

8. LIMITATIONS (be honest with prospects)

Tell prospects clearly:

"Buzz contract analysis is pattern-based AI scanning — strong at detecting known vulnerability shapes (hidden mints, blacklists, fee traps, reentrancy). It cannot detect multi-transaction state exploits, game-theory attacks, cross-protocol composability risks, or off-chain assumption bugs. For high-value listings, a formal audit from Pashov, Code4rena, or Sherlock is strongly recommended."

9. AUDIT REFERRAL (BD OPPORTUNITY)

When a contract has HIGH/CRITICAL findings or is > 3,000 lines:

Suggest formal audit partners: Pashov Audit Group, Code4rena, Sherlock
Frame as: "We recommend completing a formal audit before listing — we can fast-track your listing to front-of-queue once audit is complete"
This extends the BD relationship and filters serious projects from rug attempts

Contract Auditor Skill v1.0 — Indonesia Sprint Day 9
For Buzz BD Agent v6.2.1-bnb | SolCex Exchange
Pattern library: references/patterns.md
"Catch what humans forget to check."

contract-auditor

Files3

SKILL.mdskills/contract-auditor/SKILL.md