Mutation Testing routes the agent to the right workflow for mewt or muton, two mutation testing tools with the same interface. It covers setting up and speeding up a campaign, hunting bugs in code that survivors show to be untested, and turning results into a formal analysis report.
The skill lists the essential commands, explains what caught, uncaught, timeout and skipped outcomes mean, and ranks mutation types by severity while reminding the reader that severity ranks the mutation, not the risk.
What You Get
- A command cheat sheet for mewt init, mutate, run, status, results, print and test
- Guidance for reading outcomes such as uncaught, timeout and skipped
- A severity table for mutation types such as error replacement and condition changes
- An equivalent mutant catalog with a verification procedure
- A report template and notes on parsing output from slither-mutate, mull and dextool-mutate
Common Use Cases
- Setting up and trimming a mutation campaign that runs too long
- Reviewing which surviving mutants point to real test gaps
- Telling equivalent mutants apart from missing tests
- Using survivors in a smart contract codebase to look for bugs
What Makes It Unique
It treats surviving mutants as leads for finding bugs, not only as a coverage number, and it keeps the loading rules in one place so workflows do not load each other. It points to the tool’s own —help as the source of truth for flags, includes blockchain-specific patterns for Solidity, Move, FunC, Cairo and Solana, and can read results from other mutation tools.
Under The Hood
The skill is prose with workflow and reference files and no scripts. The agent runs mewt or muton commands in the shell, which create a config file and a SQLite database in your project, generate mutated copies of source files, and run your test suite against them. It reads and searches your code and tests and can write reports and ID lists to files. It does not call web services itself.