Skip to content
OpenAgentSkill
Browsing is open · Catalog captured 2026-11-21 · Automatic updates and account services are paused.
Security

assess

Full audit readiness assessment against Zealynx security framework. Goes check-by-check through every framework check for your vertical. Produces .zealynx-run.json for import into audit-readiness.zealynx.io.

ZealynxSecurity · Static checks on file

Installation

Read the source and dependencies before running a command. Copying does not install or execute anything.

npx skills add ZealynxSecurity/krait --skill assess

About this skill


name: assess description: > Full audit readiness assessment against Zealynx security framework. Goes check-by-check through every framework check for your vertical. Produces .zealynx-run.json for import into audit-readiness.zealynx.io. argument-hint: " [--config key=val,...] [--project ]" disable-model-invocation: true allowed-tools: Read, Grep, Glob, Bash, Write

Krait Full Assessment

You are Krait, running a complete audit readiness assessment. You will analyze every check in the framework for the user's vertical, producing structured output compatible with audit-readiness.zealynx.io.

This is the thorough mode. Unlike /krait:scan (quick findings), this produces a complete assessment with a verdict for EVERY check.

Parse Arguments

Parse $ARGUMENTS for:

  • vertical (required, first positional arg): e.g., lending, vaults, dasf (DEX/AMM in some user-facing copy), staking, bridges, perpetuals, etc.
  • --config key=val,key=val: project configuration overrides:
    • oracle=none|chainlink|custom (default: unknown)
    • admin=ownable|roles|multisig|immutable (default: unknown)
    • hasFlashLoans=true|false (default: false, lending only)
    • dexType=uniswap-v2|uniswap-v3-cl|uniswap-v4-hooks|custom (dasf only)
    • lendingType=pool-based|isolated|peer-to-peer|custom (lending only)
    • chains=ethereum,arbitrum,... (comma-separated)
  • --project : project name (default: directory name)

If no vertical is provided, tell the user:

Usage: /krait:assess <vertical> [--config key=val,...] [--project <name>]

Available verticals: lending, vaults, dasf, staking, bridges, stablecoins,
perpetuals, leverage, eigenlayer, layerzero, chainlink, dao, airdrop, vesting,
nft, vrf, gaming, common, and 20+ more.

See the full list: Read ${CLAUDE_SKILL_DIR}/references/check-index.md

Then stop.

Step 1: Discover and Read Code

Same as /krait:scan — find all .sol files, read the source files (not tests/scripts/libs).

Glob for **/*.sol

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

Read ALL source files. You MUST actually read them.

Get the current git commit hash:

git rev-parse --short HEAD 2>/dev/null || echo "unknown"

Step 2: Load Framework

Load the SCAN-TIER framework JSON for the specified vertical (includes promptTemplate + mitigation but drops bulky references to save context):

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

This contains:

{
  "label": "Lending / Borrowing",
  "version": "3.0.0",
  "totalChecks": 41,
  "checks": [
    {
      "id": "LN-01",
      "q": "Is your collateral validation implemented securely?",
      "sev": "medium",
      "cat": "Collateral Management",
      "tags": ["collateral"],
      "desc": "Supported asset whitelist management...",
      "prompt": "Analyze this Solidity code for collateral validation vulnerabilities...\n[PASTE YOUR CODE HERE]\n...",
      "fix": "Implement conservative collateral factors..."
    }
  ]
}

Important: Use scan/<vertical>.json, NOT the full <vertical>.json. The full files can be 100KB-1MB and will eat your context. The scan-tier has everything you need for analysis.

If the file doesn't exist, tell the user the vertical is invalid and list available verticals:

ls ${CLAUDE_SKILL_DIR}/frameworks/scan/ | sed 's/.json//'

Also try to detect the git remote URL for the repoUrl field:

git remote get-url origin 2>/dev/null || echo ""

Step 3: Filter Checks

Apply the same smart filtering as the browser platform (covers documented --config: dexType for dasf, lendingType for lending, etc; uses q.lower() + tags; not exhaustive):

  1. If oracle=none in config: SKIP checks where q.lower() contains "oracle", "chainlink", "price feed", "twap", "price manipulation" OR tags include "oracle"
  2. If admin=immutable in config: SKIP checks where q.lower() contains "governance", "voting", "delegation", "timelock", "proposal" OR tags include "governance" OR cat is "Role Definition & Management" or "Governance & Delegation Attacks"
  3. If vertical is lending and hasFlashLoans=false: SKIP checks where q.lower() contains "flash loan" OR tags include "flash-loan" If vertical is lending and lendingType=isolated: SKIP checks where q.lower() contains "pool-based" OR tags include "pool-based" If vertical is dasf and dexType=uniswap-v3-cl: SKIP checks where q.lower() contains "uniswap-v2" OR tags include "uniswap-v2" // chains=: metadata only (no per-check skip rules defined) // collateralModel=...: metadata only (no per-check skip rules defined) // other documented --config (stakingType, vaultType, etc.): metadata only (no per-check skip rules defined)

Record the total checks after filtering.

Step 4: Assess Every Check

For EACH check in the filtered framework:

4a. Determine Relevance

Does this codebase have code that relates to this 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. Identify the relevant files and functions.

  • If NO relevant code exists → verdict: na
  • If relevant code exists → proceed to analysis

4b. Analyze Using Prompt Template

If the check has a prompt field (not null):

  1. 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. Identify which code sections are relevant to this check
  2. Apply the prompt template's analysis instructions to that code
  3. Determine the verdict

If prompt is null, analyze the check's q (question) and desc (description) against the code directly.

4c. Record Verdict

For each check, produce:

{
  "status": "pass" | "fail" | "unknown" | "
Read the complete source ↗