fla-triton-to-gluon

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 (WGMMA / tcgen05), and scheduling (persistent kernels, warp specialization). Covers when a port is worth it, an incremental porting sequence that keeps numerical parity at every step, a Triton-to-Gluon API mapping, compile-time / autotune / smem-budget management for heavily unrolled kernels, and a pitfall checklist (proxy fences, mbarrier semantics, layout costs, bitwise-cancellation traps, NaN-poisoned OOB handling). Use when a Triton kernel is register-bound, when `num_stages` pipelining underperforms, or when Hopper/Blackwell features (TMA, TMEM, tcgen05) are needed.

Quality

86%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

81%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A thorough, actionable porting guide with an excellent validated workflow and pitfall checklist. It is dense and expert but long, and keeps all reference material inline with no bundle files to offload detail into.

Suggestions

Move the Triton→Gluon mapping table and/or the pitfall checklist into separate reference files (e.g. references/mapping.md, references/pitfalls.md) to slim the main SKILL.md and improve progressive disclosure.

Tighten the 'Environment and versions' and 'Compile time, autotune, and the smem budget' sections where some explanatory clauses could be trimmed without losing the actionable detail.

Consider replacing the pipeline skeleton pseudocode with a minimal executable cp.async or TMA example to lift actionability from mostly-executable to copy-paste ready.

Dimension	Reasoning	Score
Conciseness	Dense and expert-level with no padding or explanation of basic concepts, assuming Claude's competence; every section carries hard-won domain knowledge, though at ~250 lines it is long for a single skill file.	4 / 5
Actionability	Provides concrete imports, a Triton→Gluon API mapping table, executable benchmark/pytest commands, and specific debug flags; the pipeline skeleton is pseudocode and some API names are flagged as version-dependent, leaving minor gaps.	4 / 5
Workflow Clarity	A clear five-step porting sequence with explicit parity-test validation checkpoints ('must pass the same parity tests…before any optimization'), re-autotune feedback after each scheduling layer, and a debugging pitfall checklist for error recovery.	5 / 5
Progressive Disclosure	Well-organized into clearly headed sections with one-level-deep, clearly signaled external references (tutorial URLs, related skills); the minor gap is that substantial reference material (mapping table, pitfall checklist) is inlined rather than split into bundle files.	4 / 5
	Total	17 / 20 Passed

Description

92%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A highly specific, well-scoped description that clearly states both what the skill does and when to use it, with concrete technical trigger phrases tailored to its niche. The only minor gap is trigger-term synonym/extension coverage.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — porting a Triton kernel to Gluon, gaining explicit control over layouts/smem/async/MMA/scheduling, plus an API mapping, autotune/smem-budget management, and a pitfall checklist — giving comprehensive coverage.	5 / 5
Completeness	Explicitly answers both 'what' (a porting workflow with stated coverage) and 'when' via a concrete 'Use when…' clause with multiple trigger conditions.	5 / 5
Trigger Term Quality	Natural trigger phrases for the target developer ('register-bound', 'num_stages pipelining underperforms', 'Hopper/Blackwell features (TMA, TMEM, tcgen05)') give good coverage, though a few synonyms or extensions are missing.	4 / 5
Distinctiveness Conflict Risk	Occupies a clear niche (porting Triton kernels in fla/ops to Gluon) with distinct, technical triggers and minimal overlap with the related FLA skills.	5 / 5
	Total	19 / 20 Passed

Validation

100%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: fla-org/flash-linear-attention
Path: .agents/skills/fla-triton-to-gluon/SKILL.md
Commit: 27967b9

Reviewed: 3 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.