New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Triton Version Support

by toyaix in toyaix/triton-runner

Add, fix, or validate Triton Runner support for an exact Triton version.

MITAuto-check passedAI & LLM Engineering

Install Triton Version Support

skills CLI
$ npx skills add toyaix/triton-runner --skill triton-version-support -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install toyaix/triton-runner triton-version-support --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/toyaix/triton-runner.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/triton-version-support .claude/skills/triton-version-support && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
triton-version-support
GitHub stars
100
Token cost
~1.1k tokens
SKILL.md length
463 words
Files
2
Skills in repo
2
Repo updated
First seen
Licence
MIT

At a glance

Add, fix, or validate Triton Runner support for an exact Triton version.

  • Works in 6 steps: Confirm the exact target version → Inspect the version-sensitive surfaces → Implement the minimal version-specific… → …
  • The user names a specific Triton release such as 3.7.0
  • SKILL.md covers Workflow and Heuristics
  • Calls python and pip

What it does

Triton Version Support is an agent skill from toyaix/triton-runner. Add, fix, or validate Triton Runner support for an exact Triton version. Use when the user names a specific Triton release such as 3.7.0 or 3.8.0 and wants Codex to adapt compatibility gates, JIT shims, example docs, or regression coverage in the Triton Runner project, then verify the result with the matching Triton install and CUDA-backed tests.

Its SKILL.md is about 1.1k tokens, which your agent loads only when the skill is triggered. The skill folder holds 2 other files (for example `agents/openai.yaml`).

It sits in AI & LLM Engineering. It works with CUDA. The repository describes itself as: Multi-Level Triton Runner supporting Python, IR, PTX, AMDGCN, cubin and hasco. The licence is MIT.

When your agent uses it

  • The user names a specific Triton release such as 3.7.0
  • 3.8.0 and wants Codex to adapt compatibility gates
  • Regression coverage in the Triton Runner project
  • Then verify the result with the matching Triton install and CUDA-backed tests

Example prompts

  • “/triton-version-support”

Requirements

  • Python 3

Workflow steps

6 steps, taken from the step headings in SKILL.md.

  1. Confirm the exact target version
  2. Inspect the version-sensitive surfaces
  3. Implement the minimal version-specific change
  4. Keep examples and docs in sync
  5. Validate in escalating steps
  6. Report with concrete outcomes

What it can do on your machine

Read from SKILL.md and the folder at commit ef0efe9. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • python
    • pip

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use pip, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Triton Version Support loads about 1.1k tokens when it runs. Until then it costs about 93 tokens; SKILL.md has 463 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~93
When it runs · the whole SKILL.md, loaded when a task matches
~1.1k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from toyaix/triton-runner at commit ef0efe9, republished under its MIT licence (© toyaix). 463 words, ~1,127 tokens.

Download SKILL.mdSave it as .claude/skills/triton-version-support/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.
name
triton-version-support
description
Add, fix, or validate Triton Runner support for an exact Triton version. Use when the user names a specific Triton release such as 3.7.0 or 3.8.0 and wants Codex to adapt compatibility gates, JIT shims, example docs, or regression coverage in the Triton Runner project, then verify the result with the matching Triton install and CUDA-backed tests.

Triton Version Support

Adapt Triton Runner to the requested Triton version with the smallest defensible diff. Treat the installed Triton package as the source of truth for runtime and JIT behavior, and validate with real CUDA-backed commands before concluding the work is done.

Workflow

1. Confirm the exact target version

Read the user's requested Triton version as an exact value such as 3.7.0.

Check the active environment first:

bash
python -c "import triton; print(triton.__version__)"

If the active version does not match and the user expects end-to-end validation on that exact version, switch or install the requested version before making claims about compatibility. Use escalated execution when pip install or CUDA access is required.

2. Inspect the version-sensitive surfaces

Start by reading the files that usually gate Triton version support in this repository:

  • triton_runner/compat/version.py
  • triton_runner/compat/__init__.py
  • triton_runner/jit/versions.py
  • triton_runner/jit/api.py
  • triton_runner/jit/gluon.py
  • triton_runner/__init__.py
  • test/test.py
  • test/regression_test.py
  • examples/runner/v*/README.md

Inspect the installed Triton implementation instead of guessing API details. Prefer direct introspection:

bash
python - <<'PY'
import inspect
from triton.runtime.jit import JITFunction
print(inspect.getsource(JITFunction))
PY

Check compute_cache_key, JITFunction.run, JITFunction._do_compile, binder layout, hook signatures, and any changed launch arguments before deciding what to override.

3. Implement the minimal version-specific change

Prefer the smallest change that matches the actual Triton delta:

  • Add a new RunnerJITFunctionVx_y_z class only when the requested version materially differs from an existing one.
  • Override only run() when that is sufficient.
  • Reuse inherited helper methods when they still match the installed Triton contract.
  • Keep the code style aligned with nearby version-specific classes.
  • Update dispatch tables and support booleans after the JIT path is correct.

When extending version coverage, update the usual outer layers:

  • Support range and booleans in triton_runner/compat/version.py
  • Re-exports in triton_runner/compat/__init__.py
  • JIT dispatch in triton_runner/jit/api.py
  • Gluon dispatch in triton_runner/jit/gluon.py if applicable
  • Example directory selection via uni_triton_version
  • Regression matrix defaults in test/regression_test.py when the new version should be part of the default sweep
Show full SKILL.md (176 more words)Show less
4. Keep examples and docs in sync

If test/test.py loads commands from examples/runner/v{uni_triton_version}/README.md, create the matching example folder when introducing a new exact version. If the new release behaves like the previous one, copy the prior README first and then adjust only if needed.

Typical command:

bash
python examples/runner/python/triton/matmul.py
5. Validate in escalating steps

Run fast checks before long regressions:

bash
python -m py_compile triton_runner/jit/versions.py triton_runner/jit/api.py triton_runner/compat/version.py
python examples/runner/python/triton/matmul.py
python test/regression_test.py 3.7.0

Use escalated execution for commands that need CUDA visibility or package installation. Distinguish environment warnings from real compatibility failures; do not report unrelated pip warnings as code regressions.

6. Report with concrete outcomes

In the final summary:

  • State which Triton version was validated.
  • Name the key compatibility files changed.
  • State whether matmul.py passed.
  • State whether the targeted regression command passed.
  • Call out residual risks only when a test could not be run or a subsystem was not exercised.

Heuristics

  • Prefer primary evidence from the installed Triton package over memory.
  • Prefer minimal diffs over broad refactors.
  • Prefer exact version names such as 3.7.0 in examples and reports.
  • Prefer fixing the real failing surface instead of preemptively rewriting unrelated compatibility code.

© toyaix, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 1 other file in skills/triton-version-support of toyaix/triton-runner.

  • SKILL.md
  • agents/openai.yaml

Open the folder on GitHubat commit ef0efe9

Compare with similar skills

Triton Version Support next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Triton Version Support compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Triton Version Support this skilltoyaix/triton-runner100—~1.1kAutomated safety check: PassMIT
Esmfold2JimLiu/science-skills2274 repos~2.5kAutomated safety check: PassApache-2.0
MUSA GPU Training Optimizeropen-infra-skills/infra-skills141—~1.7kAutomated safety check: PassApache-2.0
Benchmark TuneMesh-LLM/mesh-llm3.5k—~1.6kAutomated safety check: PassApache-2.0
Cuda Kernel OptimizerKernelFlow-ops/cuda-optimized-skill213—~4.3kAutomated safety check: PassMIT
Hugging Face Local Modelshuggingface/skills11k3 repos~945Automated safety check: PassApache-2.0

Similar skills

  • Esmfold2

    JimLiu/science-skills

    Biohub ESMFold2 / ESMFold2-Fast all-atom co-folding (Candido et al.

    227 GitHub starsUsed in 4 repos~2.5k tokens
    AI & LLM EngineeringAuto-check passed

  • MUSA GPU Training Optimizer

    open-infra-skills/infra-skills

    Profiles, benchmarks and tunes AI training workloads on Moore Threads MUSA GPUs with a measurement-first process that keeps model behavior unchanged.

    141 GitHub stars~1.7k tokensUpdated 3 mo ago
    AI & LLM EngineeringAuto-check passed

  • Benchmark Tune

    Mesh-LLM/mesh-llm

    A skill your agent uses when running, debugging, interpreting, or documenting mesh-llm benchmark tune model-serving throughput trials, including choosing…

    3.5k GitHub stars~1.6k tokensUpdated today
    AI & LLM EngineeringAuto-check passed

  • Cuda Kernel Optimizer

    KernelFlow-ops/cuda-optimized-skill

    Iteratively optimize a CUDA/CUTLASS/Triton kernel only when strict on-device compilation, correctness, timing, and NCU evidence gates pass.

    213 GitHub stars~4.3k tokensUpdated 1 mo ago
    AI & LLM EngineeringAuto-check passed

  • Hugging Face Local Models

    huggingface/skills

    Official

    Finds llama.cpp-compatible GGUF models on the Hugging Face Hub, picks a quantization for your hardware and launches them with llama-cli or llama-server.

    11k GitHub starsUsed in 3 repos~945 tokens
    AI & LLM EngineeringAuto-check passed

  • Areno Develop Kernel

    inclusionAI/AReno

    Develop, optimize, debug, and validate an AReno CUDA, Triton, fused, attention, convolution, routing, or MoE operator.

    323 GitHub stars~498 tokensUpdated today
    AI & LLM EngineeringAuto-check passed

More from toyaix/triton-runner

  • Triton Generate Artifacts

    toyaix/triton-runner

    Generate Triton Runner ttir, ttgir, llir, ptx, cubin, and metadata artifacts inside examples/runner/vX.Y.Z for the active Triton version and the current machine's GPU capability.

    100 GitHub stars~471 tokensUpdated 22 days ago
    Auto-check passed

Works with

Categories

Questions about Triton Version Support

What does Triton Version Support do?

Add, fix, or validate Triton Runner support for an exact Triton version. Triton Version Support is an agent skill from toyaix/triton-runner. Add, fix, or validate Triton Runner support for an exact Triton version.

When should I use Triton Version Support?

Triton Version Support fits situations like: the user names a specific Triton release such as 3.7.0; 3.8.0 and wants Codex to adapt compatibility gates; regression coverage in the Triton Runner project; then verify the result with the matching Triton install and CUDA-backed tests.

How do I install Triton Version Support in Claude Code?

Run `npx skills add toyaix/triton-runner --skill triton-version-support -a claude-code`. Or copy the skill folder (skills/triton-version-support in toyaix/triton-runner) into .claude/skills/triton-version-support in your project. Claude Code loads it when a task matches its description.

How do I install Triton Version Support in Codex?

Run `npx skills add toyaix/triton-runner --skill triton-version-support -a codex`. Or copy the skill folder (skills/triton-version-support in toyaix/triton-runner) into .agents/skills/triton-version-support in your project. Codex loads it when a task matches its description.

Can I use Triton Version Support in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add toyaix/triton-runner --skill triton-version-support -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/triton-version-support, .gemini/skills/triton-version-support, .github/skills/triton-version-support and .opencode/skills/triton-version-support in your project.

What does Triton Version Support need to run?

Going by SKILL.md and its folder, Triton Version Support needs the command-line tools its instructions call (python and pip). Our summary lists: Python 3.

Does Triton Version Support access the network?

SKILL.md contains no URLs. Its commands use pip, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Triton Version Support safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Triton Version Support use?

Triton Version Support is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Triton Version Support use?

About 1.1k tokens (SKILL.md is roughly 4.5k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Triton Version Support?

Skills that share tags, products or a category with Triton Version Support: Esmfold2 (JimLiu/science-skills, 227 stars), MUSA GPU Training Optimizer (open-infra-skills/infra-skills, 141 stars), Benchmark Tune (Mesh-LLM/mesh-llm, 3.5k stars) and Cuda Kernel Optimizer (KernelFlow-ops/cuda-optimized-skill, 213 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Triton Version Support?

toyaix (a GitHub organization) maintains it in toyaix/triton-runner, which has 100 GitHub stars. The repository holds 2 skills in this directory. The repository was last updated on September 17, 2026.

Source: toyaix/triton-runner on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.