Cosmos Vulnerability Scanner audits custom x/ modules, IBC integrations and CosmWasm contracts against 25 core, 16 IBC, 10 EVM and 3 CosmWasm patterns, updated for SDK v0.53.
It runs a discovery agent that writes a technical inventory, spawns scanning agents in parallel for each pattern group, checks that every pattern was reported and writes each finding as its own markdown file in an output folder.
What You Get
- A CLAUDE.md in the target repository with a technical inventory and threat model
- One markdown file per finding with severity, location, vulnerable code, attack scenario and recommendation
- A report from each scanner on every pattern in its reference file, with not-applicable reasons
- A listing of all finding files in the output folder
Common Use Cases
- Auditing custom x/ modules before a chain launch
- Reviewing an IBC integration for reentrancy and token inflation
- Investigating the cause of a chain halt
- Checking CosmWasm contracts against known patterns
What Makes It Unique
The scanner is built around consensus reachability: a finding counts only when it sits on a path such as BeginBlock, EndBlock, msg_server handlers or the AnteHandler, which cuts false positives from CLI or query code. It checks SDK versions in go.mod before applying patterns, rejects rationalizations like trusting governance or an IBC counterparty, and verifies pattern counts per agent.
Under The Hood
The skill is an instruction file with a changelog and resource files of patterns. It runs subagents through the Agent tool: a discovery agent first, then scanning agents that only use Grep, Glob and Read. The main skill writes a CLAUDE.md into the root of the target repository and writes each finding as a markdown file into an output folder, by default .bughunt_cosmos. It runs no shell commands and does not fetch web pages or call APIs.