Cute Verify Agent Skill
Verify changes to the Helion CuTe backend by running both test suites, lint, and a code review subagent — then fix any issues found.
Install
one command, takes just this skill from the repositorynpx skills add https://github.com/pytorch/helion --skill cute-verifyWhat it tells the agent to use
found in the instruction textThe instruction itself
as written by the authorRun all four checks, then fix any issues:
- Lint —
./lint.sh(Ruff + Pyrefly). - Default test suite —
pytest test/ -n4 -q --tb=line. - CuTe test suite —
HELION_BACKEND=cute pytest test/ -n4 -q --tb=line. - Code review — spawn an Opus-class
general-purposesubagent with the diff (git diff HEAD~covers staged + uncommitted, or pick the right scope for the situation).
Parallelism: Run lint and the code-review subagent in parallel with the test suites is fine, but never run the two pytest suites simultaneously — both use -n4 and would put 8 workers on the GPU, causing OOMs and flaky failures. Run pytest suites sequentially.
Review subagent prompt (use subagent_type: general-purpose, model: opus):
> Review git diff HEAD~ in /data/users/jansel/helion. Focus on:
> - Correctness: wrong logic, missed edge cases, broken invariants.
> - Simplification: redundant code, layers that could collapse, premature abstractions.
> - Hacky things: ad-hoc workarounds, type-casts that hide bugs, magic constants without comments, stub APIs that aren't really plumbed.
> - Disabled / regressed coverage: tests that were deleted or @skiped without an equivalent replacement; functionality that silently lost a code path (e.g. an if branch removed without checking callers).
>
> Report concrete file:line issues. Under 400 words. Say "LGTM" plainly if clean.
If any check fails, fix the underlying issue (don't suppress it) and re-run only that check. Repeat until all four are clean.
Final report:
| Check | Result |
| ----- | ------ |
| ./lint.sh | clean / N issues fixed |
| pytest (default) | <pass>/<skip> |
| pytest (cute) | <pass>/<skip>/<xfail> |
| code review | LGTM / <N> nits applied |
Notes:
- Don't pass
HELION_AUTOTUNE_EFFORT=noneto a full suite run; it changes execution paths and breaks tests. Use it only for targeted local iteration. - Use
pytest -raif a skip count looks suspicious — it prints skip reasons.
Other skills for the same job
different authors, same section of the catalogue
You are an expert AI-powered code review specialist combining automated static analysis, intelligent pattern recognition, and modern DevOps practices. Leverage AI tools (GitHub Copilot, Qodo, GPT-5, C
Write idiomatic Ruby code with metaprogramming, Rails patterns, and performance optimization. Specializes in Ruby on Rails, gem development, and testing frameworks. Use PROACTIVELY for Ruby refactoring, optimization, or complex Ruby features.
Use when working with tdd workflows tdd refactor
Set up Husky pre-commit hooks with lint-staged (Prettier), type checking, and tests in the current repo. Use when user wants to add pre-commit hooks, set up Husky, configure lint-staged, or add commit-time formatting/typechecking/testing.
AI-assisted pair programming with multiple modes (driver/navigator/switch), real-time verification, quality monitoring, and comprehensive testing. Supports TDD, debugging, refactoring, and learning sessions. Features automatic role switching, continuous code review, security scanning, and performance optimization with truth-score verification.
Prepares codebases for security review using Trail of Bits' checklist. Helps set review goals, runs static analysis tools, increases test coverage, removes dead code, ensures accessibility, and generates documentation (flowcharts, user stories, inline comments).
Refactor high-complexity React components in Dify frontend. Use when `pnpm analyze-component --json` shows complexity > 50 or lineCount > 300, when the user asks for code splitting, hook extraction, or complexity reduction, or when `pnpm analyze-component` warns to refactor before testing; avoid for simple/well-structured components, third-party wrappers, or when the user explicitly wants testing without refactoring.
Write modern C# code with advanced features like records, pattern matching, and async/await. Optimizes .NET applications, implements enterprise patterns, and ensures comprehensive testing. Use PROACTIVELY for C# refactoring, performance optimization, or complex .NET solutions.
How to use it
Take pytorch/cute-verify from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.