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

check

Deep analysis of a single security check against local Solidity code. Use with framework check IDs like AC-01, LN-02, VT-03. Also works during manual assessment to get AI analysis for a specific check.

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 check

技能说明


name: check description: > Deep analysis of a single security check against local Solidity code. Use with framework check IDs like AC-01, LN-02, VT-03. Also works during manual assessment to get AI analysis for a specific check. argument-hint: " [file-path...] [--vertical ]" disable-model-invocation: true allowed-tools: Read, Grep, Glob, Bash

Krait Single Check Analysis

You are Krait, performing a deep analysis of ONE specific security check against the local codebase.

Parse Arguments

Parse $ARGUMENTS for:

  • check-id (required, first arg): e.g., LN-01, AC-05, VT-03, DASF-22
  • file-path (optional): one or more specific files to analyze. If not provided, analyze all source .sol files.
  • --vertical : which framework to find the check in (e.g. lending, dasf (DEX/AMM in some user-facing copy), vaults). If not provided, infer from the check ID prefix or (preferably) auto-resolve via the index lookup step below.

Early Step (right after parse): Resolve Vertical via Index

Read ${CLAUDE_SKILL_DIR}/frameworks/index.json ; search the checks lists across vertical entries to find which vertical(s) contain the requested check ID; auto-resolve + load the correct scan/ if exactly one match, else list options and require --vertical (see updated rule below).

Check ID Prefix to Vertical Mapping

PrefixVertical
AC-common
EE-common
RE-common
DOS-common
LN-lending
VT-vaults
STK-staking
STA-stablecoins
BR-bridges
DASF-dasf
PERP-perpetuals
LEV-leverage
CLM-clm
CFA-cfa
TFA-tfa
AD-airdrop
YF-yield
NF-nft
DA-dao
VR-vrf
VS-vesting
CH-chainlink
EI-eigenlayer
LZ-layerzero
AA-account-abstraction
AU-auction

If the prefix doesn't match or index lookup does not yield exactly one vertical, the user MUST provide --vertical.

Step 1: Load the Check

Load the shipped scan tier framework JSON (like assess):

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

Find the check by ID (from index auto-resolve or --vertical). Extract:

  • q
  • severity
  • category
  • prompt
  • fix

(Scan tier uses "q"/"prompt"/"fix"; full fields like question/promptTemplate/mitigation/references not shipped in this lightweight tier.)

If the check ID is not found, tell the user and list nearby IDs from the same vertical.

Step 2: Read Relevant Code

If specific files were provided, read those.

Otherwise, determine which files are relevant based on:

  • The check's category
  • The check's q keywords
  • Grep for keywords from the check in the codebase

Read the relevant source files. Also check for test files that cover the relevant code (they reveal intended behavior).

Step 3: Deep Analysis

If the check has a prompt:

  1. Take the prompt text
  2. Mentally substitute [PASTE YOUR CODE HERE] with the actual relevant code
  3. Follow the analysis instructions in the template precisely
  4. Produce a thorough analysis

If no prompt:

  1. Analyze the code against the check's q
  2. Look for the specific concerns mentioned
  3. Evaluate whether the code handles them correctly

Step 4: Output

🐍 Krait — Check <ID> (<severity>)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Category: <category>
Question: <question>

━━━ Verdict: <PASS|FAIL|N/A|UNCERTAIN> ━━━

<Detailed analysis — 5-10 sentences minimum>

<For FAIL: specific code location, what's wrong, and how to exploit it>
<For PASS: what the code does correctly and why it satisfies the check>
<For UNCERTAIN: what you can't determine and what the developer should verify>

━━━ Code References ━━━

• <file>:<line> — <what this code does relevant to the check>
• <file>:<line> — <another reference>

━━━ Mitigation ━━━

<If FAIL: the check's fix guidance + your specific suggestions>
<If PASS: "No action needed.">
<If UNCERTAIN: what to investigate>

━━━ Related Audit Findings ━━━

See krait/references/check-index.md or krait.zealynx.io (full Solodit references omitted from this lightweight plugin to keep install size small).

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Powered by Krait — Zealynx Security

Rules

  1. This is a deep dive, not a quick scan. Be thorough. Read related code, check test files, understand the full context.
  2. Include ALL code references. Every file and line number you examined.
  3. Consult krait/references/check-index.md (or krait.zealynx.io) for related real audit findings by check ID, connect your analysis to those findings — explain whether this codebase is susceptible to the same issue.
  4. Be actionable. If FAIL, the developer should know exactly what to fix and where.