CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

at-dispatch-v2

Convert PyTorch AT_DISPATCH macros to AT_DISPATCH_V2 format in ATen C++ code. Use when porting AT_DISPATCH_ALL_TYPES_AND*, AT_DISPATCH_FLOATING_TYPES*, or other dispatch macros to the new v2 API. For ATen kernel files, CUDA kernels, and native operator implementations.

Quality

83%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

71%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.

The body is highly actionable with complete before/after transformations and a clear step sequence, but it is roughly twice as long as needed: the same conversion pattern is demonstrated many times over, and a large lookup catalog is inlined instead of split into a reference file. Consolidating the examples and moving the type-group/pattern catalog to references/ would raise both conciseness and progressive disclosure without losing actionability.

Suggestions

Remove the 'Quick reference', 'Benefits of AT_DISPATCH_V2', and redundant 'Common patterns'/'Edge cases' sections; one canonical before/after example (Step 5) plus the multi-line lambda example (Step 6) already demonstrates the full transformation.

Move the 'Type group reference' table and remaining pattern variants into a references/ file (e.g. references/patterns.md) linked from a short 'Pattern catalog' section, keeping SKILL.md as the procedural core.

Add a feedback loop to the Workflow section (e.g. re-check each converted macro against the Step 7 checklist before moving to the next) so the verification checkpoint drives iteration rather than sitting at the end.

DimensionReasoningScore

Conciseness

All content is on-topic and domain-specific, but there is notable redundancy: the 'Quick reference' section duplicates the Step 5 example, 'Common patterns' and 'Edge cases' re-illustrate the same transformation five-plus times, and 'Benefits of AT_DISPATCH_V2' argues for the API rather than instructing. Fits anchor 3 ('mostly efficient but could be tightened'), not 4 because the duplication exceeds minor trimming, and not 2 because there is no generic filler or explanation of concepts Claude already knows.

3 / 5

Actionability

Fully executable, copy-paste-ready before/after code for every pattern, concrete step-by-step instructions, explicit include placement, and a verification checklist. Specific examples cover the common cases (ALL_TYPES_ANDn, FLOATING_TYPES_ANDn, COMPLEX variants, no-capture lambdas), matching anchor 5.

5 / 5

Workflow Clarity

Steps 1-7 are clearly sequenced, reinforced by a summary Workflow section and a Step 7 verification checklist, so the batch-operation validation cap does not apply. Not 5 because there is no fix-and-retry feedback loop and 'Do NOT compile or test' precludes executable validation; not 3 because checkpoints are explicit, not missing.

4 / 5

Progressive Disclosure

Headers make the 300-line file navigable, but roughly half of it (type group reference, repeated pattern catalog, edge cases) is lookup material that belongs in a one-level-deep reference file rather than in SKILL.md. Matches anchor 3's example of a large inline reference that could be separated, not 4 given the volume of inline reference content and absence of any bundle files.

3 / 5

Total

15

/

20

Passed

Description

87%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 strong description: it states what the skill does, when to use it, and the exact macro families involved, using natural domain vocabulary in third person. The only gap is that a few trigger terms the body itself recognizes (e.g. 'Dispatch_v2.h', 'dispatch v2') are missing from the description itself.

DimensionReasoningScore

Specificity

Names the domain (PyTorch ATen C++) and concrete conversion actions, enumerating specific macro families like 'AT_DISPATCH_ALL_TYPES_AND*' and 'AT_DISPATCH_FLOATING_TYPES*'. Not 5 because the actions are variations of a single conversion task rather than multiple distinct capabilities; not 3 because it goes well beyond naming 1-2 bare actions by listing the specific macro families covered.

4 / 5

Completeness

Explicitly answers both what ('Convert PyTorch AT_DISPATCH macros to AT_DISPATCH_V2 format in ATen C++ code') and when ('Use when porting AT_DISPATCH_ALL_TYPES_AND*, AT_DISPATCH_FLOATING_TYPES*, or other dispatch macros to the new v2 API') with concrete trigger phrases, in third person. This matches the anchor-5 example structure directly.

5 / 5

Trigger Term Quality

Includes terms a user would naturally say for this task: 'AT_DISPATCH_ALL_TYPES_AND*', 'dispatch macros', 'the new v2 API', 'ATen kernel files', 'CUDA kernels'. Not 5 because natural triggers the body itself identifies ('Dispatch_v2.h', 'dispatch v2', 'macro conversion') are absent from the description; not 3 because coverage of the domain's actual vocabulary is strong, not partial.

4 / 5

Distinctiveness Conflict Risk

The AT_DISPATCH/AT_DISPATCH_V2 macro niche in PyTorch's ATen codebase is extremely specific with distinct triggers; no realistic overlap with other skills. Clearly matches the anchor-5 'clear niche with distinct triggers; minimal conflict risk'.

5 / 5

Total

18

/

20

Passed

Validation

100%

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

Microck/ordinary-claude-skills

Path

skills_all/at-dispatch-v2/SKILL.md

Reviewed

Table of Contents

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.