New: 97,000+ 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

Fla Correctness Coverage

by fla-org in fla-org/flash-linear-attention

Guidelines for kernel correctness testing and coverage in fla/ops/ and related modules, including common Triton grid/addressing pitfalls.

MITAuto-check passed

Install Fla Correctness Coverage

skills CLI
$ npx skills add fla-org/flash-linear-attention --skill fla-correctness-coverage -a claude-code

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

GitHub CLI
$ gh skill install fla-org/flash-linear-attention fla-correctness-coverage --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/fla-org/flash-linear-attention.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.agents/skills/fla-correctness-coverage .claude/skills/fla-correctness-coverage && 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
fla-correctness-coverage
GitHub stars
5.8k
Token cost
~1.2k tokens
SKILL.md length
521 words
Files
5 (incl. references)
Skills in repo
9
Repo updated
First seen
Licence
MIT

At a glance

Guidelines for kernel correctness testing and coverage in fla/ops/ and related modules, including common Triton grid/addressing pitfalls.

  • Works in 4 steps: List the current coverage matrix for the… → Compare against the axes below. → Add tests for missing combinations that… → …
  • SKILL.md covers Workflow, Public reference docs, Coverage axes and Kernel implementation safety…, plus 4 more sections
  • Calls pytest and python

What it does

Fla Correctness Coverage is an agent skill from fla-org/flash-linear-attention. Guidelines for kernel correctness testing and coverage in fla/ops/ and related modules, including common Triton grid/addressing pitfalls. Helps decide what tests to add or run before an MR.

Its SKILL.md is about 1.2k tokens, which your agent loads only when the skill is triggered. The skill folder holds 5 other files, including reference files (for example `references/cp.md`, `references/delta-rule.md` and `references/generalized-delta-rule.md`).

The repository describes itself as: 🚀 Efficient implementations for emerging model architectures. The licence is MIT.

Example prompts

  • “Use the fla-correctness-coverage skill to guideline for kernel correctness testing and coverage in fla/ops/ and related modules, including common…”
  • “/fla-correctness-coverage”

Requirements

  • Python 3

Workflow steps

4 steps, taken from the first numbered list in SKILL.md.

  1. List the current coverage matrix for the op you are touching.
  2. Compare against the axes below.
  3. Add tests for missing combinations that are reachable by user code.
  4. Run the relevant tests and make sure they pass.

What it can do on your machine

Read from SKILL.md and the folder at commit b8ff848. 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:

    • pytest
    • python

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

  • Network

    No URLs in SKILL.md.

    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

Fla Correctness Coverage loads about 1.2k tokens when it runs, and up to ~1.2k if it reads all its reference files. Until then it costs about 54 tokens; SKILL.md has 521 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~54
When it runs · the whole SKILL.md, loaded when a task matches
~1.2k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~1.2k

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 fla-org/flash-linear-attention at commit b8ff848, republished under its MIT licence (© fla-org). 521 words, ~1,153 tokens.

Download SKILL.mdSave it as .claude/skills/fla-correctness-coverage/SKILL.md (or your agent's skills folder). This skill also uses 4 other files; get the full folder from GitHub.
name
fla-correctness-coverage
description
Guidelines for kernel correctness testing and coverage in fla/ops/** and related modules, including common Triton grid/addressing pitfalls. Helps decide what tests to add or run before an MR.

FLA Correctness & Coverage Skill

Use this skill when adding or modifying a kernel in fla/ops/ (e.g., KDA, GDN, GLA, DeltaNet, NSA, etc.) and you need to verify correctness or close a coverage gap.

Workflow

  1. List the current coverage matrix for the op you are touching.
  2. Compare against the axes below.
  3. Add tests for missing combinations that are reachable by user code.
  4. Run the relevant tests and make sure they pass.

Public reference docs

When a task needs operator math or protocol details, read only the relevant reference file:

  • references/cp.md — context parallelism for linear attention, including KDA/GDN CP formulation.
  • references/delta-rule.md — Delta Rule operator background.
  • references/generalized-delta-rule.md — Generalized Delta Rule operator background.
  • references/simple-gla.md — Simple GLA operator background.

Do not load every reference by default; use these only when the touched code or test depends on that operator's math or distributed protocol.

Coverage axes

For each kernel, check coverage across these dimensions:

AxisValues to cover
Sequence layoutdense, variable-length (varlen)
Directionforward, backward
Gate modesafe gate, non-safe gate (if applicable)
Beta moderaw beta, post-sigmoid beta (if applicable)
QK normalizationwith L2 norm, without L2 norm
Stateinitial state, final state (if the op supports state passing)
GVAgrouped value attention (GVA) enabled vs disabled
Head dimensionsD != Dv (different qk and v head dims)
Backend verifierreference implementation, torch.autograd.gradcheck, and backend-specific sanity checks

Kernel implementation safety checks

Before adding or changing a Triton kernel, check these implementation details in addition to numerical tests:

  • Treat program IDs and grid-derived values as potentially narrow. On NVIDIA, non-first grid dimensions may be narrow; on AMD, Ascend, or other non-NVIDIA backends, every grid dimension may be narrow. Cast to tl.int64 before using them in address arithmetic.
  • Keep tensor address arithmetic in tl.int64, including block bases, strides, varlen sequence offsets, head offsets, and element offsets. Do not rely on int16 or int32 overflow behavior.
  • Do not introduce new tl.make_block_ptr use. Triton marks it deprecated; use TensorDescriptor / tl.make_tensor_descriptor when descriptor semantics are needed, or explicit tl.load / tl.store pointer arithmetic following an existing validated kernel pattern.
  • If a change touches grid shape, program-id mapping, varlen offsets, or pointer math, run a shape that exercises the changed path on NVIDIA and any supported non-NVIDIA backend, or add a precise verifier/skip for unsupported platforms.
Show full SKILL.md (143 more words)Show less

Code style constraints

  • Use fla.utils.device and fla.utils.device_platform in tests instead of adding new hard-coded device strings.
  • Use IS_NVIDIA, IS_NVIDIA_HOPPER, IS_NVIDIA_BLACKWELL, IS_AMD, and IS_INTEL from fla.utils for platform-specific skips or branches.
  • Do not add new direct torch.cuda platform checks in correctness tests. If no existing helper covers the condition, add a small helper in fla.utils first.

Default open-source test paths

Use these paths when looking for existing tests or deciding where to add new ones:

  • tests/ops/test_kda.py — KDA kernel tests
  • tests/context_parallel/ — context-parallel variants (e.g., test_cp_kda.py, test_cp_gdn.py)
  • tests/models/test_modeling_kda.py — end-to-end model tests for KDA

Adapt the path to the specific op you are working on (replace kda with gdn, gla, nsa, delta, etc.).

What NOT to put in this skill

  • Internal-only test paths, local machine paths, private model names, and private workload identifiers.
  • The open-source skill only points to public tests and public operator docs.

Running tests

bash
# Single op test
pytest tests/ops/test_kda.py -v

# Context parallel tests for the same op
pytest tests/context_parallel/test_cp_kda.py -v

# Model-level test
pytest tests/models/test_modeling_kda.py -v

# All dependent tests (see fla-mr-readiness skill)
python scripts/find_dependent_tests.py <changed_files>

© fla-org, 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 4 other files (references) in .agents/skills/fla-correctness-coverage of fla-org/flash-linear-attention.

  • SKILL.md
  • references/cp.md
  • references/delta-rule.md
  • references/generalized-delta-rule.md
  • references/simple-gla.md

Open the folder on GitHubat commit b8ff848

Compare with similar skills

Fla Correctness Coverage 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.

Fla Correctness Coverage compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Fla Correctness Coverage this skillfla-org/flash-linear-attention5.8k—~1.2kAutomated safety check: PassMIT
Review Op Test CoverageCVCUDA/CV-CUDA2.7k—~264Automated safety check: PassCustom licence
Coveragealirezarezvani/claude-skills28k2 repos~669Automated safety check: PassMIT
Test Coveragethedaviddias/Front-End-Checklist74k—~407Automated safety check: PassMIT
Review Op Bench CoverageCVCUDA/CV-CUDA2.7k—~274Automated safety check: PassCustom licence
Analyzing CoverageTriliumNext/Trilium38k—~2kAutomated safety check: PassAGPL-3.0

Similar skills

  • Review Op Test Coverage

    CVCUDA/CV-CUDA

    Review a CV-CUDA operator's test coverage, including C++ correctness, required cross-layout parity, correctness rigor, and the Python API surface.

    2.7k GitHub stars~264 tokensUpdated 20 days ago
    Testing & QAAuto-check passed

  • Coverage

    alirezarezvani/claude-skills

    Analyze test coverage gaps. An agent skill from alirezarezvani/claude-skills.

    28k GitHub starsUsed in 2 repos~669 tokens
    Testing & QAAuto-check passed

  • Test Coverage

    thedaviddias/Front-End-Checklist

    A skill your agent uses when reviewing CI coverage, automated checks, or test strategy related to Maintain test coverage thresholds.

    74k GitHub stars~407 tokensUpdated yesterday
    Testing & QAAuto-check passed

  • Review Op Bench Coverage

    CVCUDA/CV-CUDA

    Review a CV-CUDA operator's BENCHMARK coverage — drivers, layout axis, baselines, the basic-tier floor, row counts, and coverage statistics.

    2.7k GitHub stars~274 tokensUpdated 20 days ago
    Data & AnalyticsAuto-check passed

  • Analyzing Coverage

    TriliumNext/Trilium

    A skill your agent uses when measuring or chasing Vitest/v8 code coverage in the Trilium monorepo — "what's below 100%?", "which files need tests?", "what lines of X are uncovered?", "take <area to…

    38k GitHub stars~2k tokensUpdated today
    Testing & QAAuto-check passed

  • Test Coverage Improver

    openai/openai-agents-python

    Official

    Measure Python SDK coverage or address measured coverage gaps.

    30k GitHub stars~687 tokensUpdated today
    Testing & QAAuto-check passed

More from fla-org/flash-linear-attention

All 9 skills in this repo

  • Fla Ascend Performance

    fla-org/flash-linear-attention

    Guidelines for Ascend NPU kernel / Triton-Ascend backend performance work in the FLA repo.

    5.8k GitHub stars~5.6k tokensUpdated yesterday
    Auto-check passed

  • Fla Optimization Loop

    fla-org/flash-linear-attention

    Disciplined, reproducible loop for making an FLA kernel faster (Triton, Gluon, TileLang, CuTe) without ever breaking or gaming correctness.

    5.8k GitHub stars~2.6k tokensUpdated yesterday
    Auto-check passed

  • Fla Triton To Gluon

    fla-org/flash-linear-attention

    Workflow for porting an existing Triton kernel in fla/ops/ to Gluon (triton.experimental.gluon) to gain explicit control over tensor layouts, shared memory, async data movement (cp.async / TMA), MMA…

    5.8k GitHub stars~4.2k tokensUpdated yesterday
    Auto-check passed

  • Fla Design Coverage

    fla-org/flash-linear-attention

    Contract-first design and coverage discipline for FLA kernel and numerical changes.

    5.8k GitHub stars~3.3k tokensUpdated yesterday
    Auto-check passed

  • Fla Dispatch Backends

    fla-org/flash-linear-attention

    Workflow for FLA backend dispatch decorators and backend implementations.

    5.8k GitHub stars~1.1k tokensUpdated yesterday
    Auto-check passed

  • Fla Kda

    fla-org/flash-linear-attention

    FLA KDA kernel workflow and public technical notes. An agent skill from fla-org/flash-linear-attention.

    5.8k GitHub stars~1.3k tokensUpdated yesterday
    Auto-check passed

Questions about Fla Correctness Coverage

What does Fla Correctness Coverage do?

Guidelines for kernel correctness testing and coverage in fla/ops/ and related modules, including common Triton grid/addressing pitfalls. Fla Correctness Coverage is an agent skill from fla-org/flash-linear-attention. Guidelines for kernel correctness testing and coverage in fla/ops/ and related modules, including common Triton grid/addressing pitfalls.

How do I install Fla Correctness Coverage in Claude Code?

Run `npx skills add fla-org/flash-linear-attention --skill fla-correctness-coverage -a claude-code`. Or copy the skill folder (.agents/skills/fla-correctness-coverage in fla-org/flash-linear-attention) into .claude/skills/fla-correctness-coverage in your project. Claude Code loads it when a task matches its description.

How do I install Fla Correctness Coverage in Codex?

Run `npx skills add fla-org/flash-linear-attention --skill fla-correctness-coverage -a codex`. Or copy the skill folder (.agents/skills/fla-correctness-coverage in fla-org/flash-linear-attention) into .agents/skills/fla-correctness-coverage in your project. Codex loads it when a task matches its description.

Can I use Fla Correctness Coverage 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 fla-org/flash-linear-attention --skill fla-correctness-coverage -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/fla-correctness-coverage, .gemini/skills/fla-correctness-coverage, .github/skills/fla-correctness-coverage and .opencode/skills/fla-correctness-coverage in your project.

What does Fla Correctness Coverage need to run?

Going by SKILL.md and its folder, Fla Correctness Coverage needs the command-line tools its instructions call (pytest and python). Our summary lists: Python 3.

Does Fla Correctness Coverage access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Fla Correctness Coverage 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 Fla Correctness Coverage use?

Fla Correctness Coverage 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 Fla Correctness Coverage use?

About 1.2k tokens (SKILL.md is roughly 4.6k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 41 tokens, read only when the agent opens those files.

What are the alternatives to Fla Correctness Coverage?

Skills that share tags, products or a category with Fla Correctness Coverage: Review Op Test Coverage (CVCUDA/CV-CUDA, 2.7k stars), Coverage (alirezarezvani/claude-skills, 28k stars), Test Coverage (thedaviddias/Front-End-Checklist, 74k stars) and Review Op Bench Coverage (CVCUDA/CV-CUDA, 2.7k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Fla Correctness Coverage?

fla-org (a GitHub organization) maintains it in fla-org/flash-linear-attention, which has 5,828 GitHub stars. The repository holds 9 skills in this directory. The repository was last updated on October 6, 2026.

Source: fla-org/flash-linear-attention on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.