- Home
- Skills
- Web3 & Crypto
- ContractGuard Lite
Works with the AI Agents you already use
ContractGuard Lite
A fast first-pass security scan for Solidity contracts, with every finding explained in plain English and paired with a concrete fix.
$9.99
ContractGuard Lite
Example session with this skill installed
Scan contracts/Vault.sol for common vulnerabilities before we deploy. I want the report and the JSON.
- Read your context and instructions
- Compiled the contractguard lite
Risk rating CRITICAL, 2 Critical, 0 High, 2 Medium, 1 Low. Top 3: Critical reentrancy at Vault.sol:29, Critical missing access control at Vault.sol:38, Medium missing zero address validation at Vault.sol:35. Full report and findings.json with evidence lines, attack narratives for High and Critical, and concrete fixes.
Connects securely to your tools. The creator never sees your data.
About this skill
Point the skill at a .sol file or a project. Get 20 canonical vulnerability checks, severity rated findings with exact file and line locations, plain English explanations, attack narratives for the serious ones, and small fixes the developer can apply. Markdown report plus JSON. No compiler, no API keys, no network.
Why it matters A professional smart contract audit costs five figures and takes weeks to schedule. Indie teams, hackathon builders, and students ship with zero review because the audit is out of reach. The free scanners that do exist dump warnings written for security researchers, which teaches the developer nothing. ContractGuard Lite is the readable first pass. It finds the classic mistakes, explains what an attacker actually does, and shows the smallest practical fix.
What it does Give it a contract file, a project folder, or pasted source. It returns a risk rating, counts by severity, the top 3 fixes first, then findings by severity, then blind spots and the disclaimer. Example from the bundled sample vault:
- Risk rating: CRITICAL, 2 Critical, 0 High, 2 Medium, 1 Low
- Top 3 to fix first: Critical reentrancy at Vault.sol:29, Critical missing access control at Vault.sol:38, Medium missing zero address validation at Vault.sol:35 Then each finding opens with the evidence line, the plain English explanation, an attack narrative for High and Critical issues, and a diff where the change is local, or a clearly labelled design change where it is not.
Twenty canonical checks Reentrancy, unchecked return values, pre-0.8 overflow, tx.origin, unsafe delegatecall, unprotected selfdestruct, missing access control, unprotected initializers, timestamp dependence, weak randomness, unbounded loops, push payments, floating pragmas, outdated compilers, zero addresses, locked value, centralization, the ERC-20 approve race, shadowing, and missing input validation.
What you get
- Exact locations: every finding names the file and line; function scoped findings also name the function and contract, with the evidence line so the reader can verify the claim in seconds.
- Severity with a rubric: Critical, High, Medium, Low, and Informational, calibrated in references/severity.md so ratings do not drift between runs.
- Attack narratives for High and Critical: what the attacker does, in order, and what they walk away with.
- Concrete fixes: a source specific diff where the change is local, or a clearly labelled design change where it is not, plus one sentence on why it works.
- A report built to decide: executive summary first, counts by severity, top 3 fixes, then findings by severity, then blind spots and the disclaimer.
- A JSON artifact for tooling: stable fields, a fail on gate, and exit code 3 for an incomplete scan, so CI cannot treat a parse error as a pass.
- No setup: pure Python standard library. No solc, no Slither, no Forge, no network, no API keys.
How it works
- The skill collects the source and confirms the scope.
- The bundled scanner lexes and parses the Solidity, runs the 20 checks, deduplicates, and ranks the findings.
- It writes findings.json and report.md from the same result, so the two cannot drift.
- The agent reads the JSON, checks the High and Critical evidence, and gives a short handoff: risk, counts, top 3, artifact paths, blind spots, and the professional audit disclaimer. The scanner is a token level structural parser, not a compiler. It does not build a full call graph or a storage layout. That limit is stated in the report, not hidden.
Requirements An agent environment that can run shell commands with Python 3.9 or newer. Standard library only. No API keys, no network calls, no Solidity toolchain.
Honest limits
- It is a pattern scanner, not a full audit. It cannot see economic design, oracle manipulation, governance capture, offchain components, cross-function reentrancy, interface calls, or novel logic bugs.
- It does not fetch verified source from a block explorer. Version 1 is source in, report out.
- It does not compile or deploy anything, and it does not run Slither, fuzzing, or formal verification.
- False positives are possible. The report shows the evidence line so a developer can check it. A confirmed false positive should be reported, not silently removed.
- False negatives are possible. A clean scan means no findings within these 20 checks, not a guarantee of safety.
- The sample contracts are fictional fixtures, not deployed code.
- Not for Vyper, Cairo, offchain code, economic design review, or a substitute for a professional audit.
What is in the package SKILL.md, README.md, LICENSE, three references (checks, severity, report format), the scanner (scan, report, seven supporting modules, and the check package), a complete sample vault with its generated report and JSON, a clean contract, previews, and an offline regression suite with one vulnerable corpus contract per check. Test suite: 52 tests passing.
It is a pattern scanner, not a full audit. It cannot see economic design, oracle manipulation, governance capture, offchain components, cross-function reentrancy, interface calls, or novel logic bugs. It does not compile or deploy anything, and it does not run Slither, fuzzing, or formal verification. False positives and false negatives are possible. A clean scan means no findings within these 20 checks, not a guarantee of safety. Not for Vyper, Cairo, offchain code, economic design review, or a substitute for a professional audit.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 23 days ago
Requires