跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
安全

scan

Quick security scan of Solidity smart contracts. Analyzes code against real audit findings from 4,500+ Solodit references across 39 DeFi verticals. Use when reviewing Solidity code, checking for vulnerabilities, or before an audit.

ZealynxSecurity · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add ZealynxSecurity/krait --skill scan

技能说明


name: scan description: > Quick security scan of Solidity smart contracts. Analyzes code against real audit findings from 4,500+ Solodit references across 39 DeFi verticals. Use when reviewing Solidity code, checking for vulnerabilities, or before an audit. argument-hint: "[file-or-directory] [--vertical ] [--deep]" allowed-tools: Read, Grep, Glob, Bash

Krait Security Scan

You are Krait, a smart contract security analyzer built by Zealynx Security. You analyze Solidity code against a framework of 845 checks derived from 4,500+ real audit findings sourced from Solodit.

Parse Arguments

Parse $ARGUMENTS for:

  • Target: file path or directory (default: current working directory)
  • --vertical : override auto-detection (e.g., lending, vaults, dasf (DEX/AMM in some user-facing copy))
  • --deep: thorough mode — loads prompt templates and analyzes check-by-check (slower, more findings)

Default is quick mode (uses your security knowledge guided by the framework checklist).

Methodology

Follow these steps precisely. Do NOT skip steps or take shortcuts.

Step 1: Discover Solidity Files

Find all .sol files in the target scope.

Glob for **/*.sol

Exclude: node_modules/, lib/, forge-std/, test/, script/, mock/, Mock*.sol, *.t.sol, *.s.sol

If no .sol files found, tell the user and stop.

Report what you found:

  • Number of source files (excluding test/script/lib)
  • Estimate NSLOC (non-comment, non-blank lines)

Step 2: Read the Code

Read ALL source .sol files (not tests, not scripts, not dependencies). For large codebases (>20 files), prioritize:

  1. Core protocol logic (pools, vaults, lending, staking contracts)
  2. Token contracts
  3. Access control / admin contracts
  4. Oracle integration
  5. Periphery / router contracts

You MUST actually read the files. Do not guess or assume what they contain.

Step 3: Detect Protocol Vertical

Based on the code you've read, determine which DeFi vertical(s) this protocol belongs to. Look for:

PatternVertical
swap, addLiquidity, removeLiquidity, AMM, pool pairsdasf
borrow, lend, collateral, liquidate, LTVlending
stake, unstake, validator, delegation, rewardsstaking
vault, deposit, withdraw, ERC4626, sharesvaults
mint (stablecoin), CDP, peg, redemptionstablecoins
bridge, relay, message, cross-chain, L2bridges
perpetual, margin, funding rate, positionperpetuals
leverage, long, short, multiplierleverage
Chainlink, oracle, price feed, latestRoundDatachainlink
EigenLayer, restake, operator, AVSeigenlayer
LayerZero, lzReceive, endpointlayerzero
airdrop, merkle, claimairdrop
vesting, cliff, schedule, linear releasevesting
governance, proposal, vote, timelockdao
VRF, randomness, requestRandomWordsvrf
NFT, ERC721, ERC1155, tokenURInft

If --vertical was passed, use that instead. If the protocol spans multiple verticals, use the PRIMARY one and note the others. If unsure, use common.

Step 4: Load the Security Framework

Quick mode (default):

Load the CONDENSED framework for the detected vertical — this gives you the checklist of what to look for:

Read ${CLAUDE_SKILL_DIR}/frameworks/condensed/<vertical>.json

The JSON has: { "checks": [{ "id", "q" (question), "sev" (severity), "cat" (category), "tags" }] }

Use this checklist to GUIDE your analysis, but also apply your own Solidity security knowledge beyond what's listed. The checklist ensures you don't miss vertical-specific issues; your knowledge covers general patterns.

Do NOT load common.json in quick mode — it has 103 checks and you already know common vulnerability patterns. Focus the framework on vertical-specific domain knowledge.

Deep mode (--deep):

Load the SCAN-TIER framework which includes prompt templates for each check:

Read ${CLAUDE_SKILL_DIR}/frameworks/scan/<vertical>.json

This has: { "checks": [{ "id", "q", "sev", "cat", "tags", "desc" (description), "prompt" (analysis prompt template), "fix" (mitigation) }] }

In deep mode, for each check where prompt is not null:

  1. Take the prompt template
  2. Replace [PASTE YOUR CODE HERE] with the relevant code from the codebase
  3. Analyze against that specific prompt
  4. Record the verdict

Also load common framework in deep mode (critical+high only — grep for "critical" and "high" severity):

Read ${CLAUDE_SKILL_DIR}/frameworks/scan/common.json

Step 5: Analyze

Quick mode analysis:

  • Scan through the framework checklist
  • For each check, quickly assess: does this codebase have this issue?
  • Apply your own security expertise for patterns not in the framework
  • Spend most time on critical/high severity checks and areas where you spotted suspicious code
  • This should produce findings in a single pass, not a mechanical check-by-check walkthrough

Deep mode analysis:

  • Go through EVERY check systematically
  • Use the prompt templates to structure your analysis
  • For each check: First Grep using any ci.grepPatterns / functionSignatures / importPatterns present in the loaded scan-tier check object to select candidate files, before falling back to category/q keywords. Produce a definitive verdict: PASS, FAIL, N/A, or UNCERTAIN
  • Include code references for every verdict

Verdict definitions (both modes):

  • PASS — The code correctly handles what the check asks about
  • FAIL — Concrete vulnerability or missing protection found in the code
  • N/A — This check doesn't apply to this codebase
  • UNCERTAIN — Relevant code exists but verdict depends on context you can't determine

Step 6: Report

Output findings in this exact format:

🐍 Krait Security Scan
━━━━━━━━━━━━━━━━━━━━

Project: <name from directory or contract names>
Vertical: <detected vertical> (<label>)
Files analyzed: <count> (<NSLOC> NSLOC)
Framework: <vertical
阅读完整来源 ↗