AF dataset-producer
Produce complete, publish-ready AI/ML datasets in HuggingFace format (parquet shards + README.md with YAML frontmatter + dataset card + provenance script). Use whenever the user wants to create, build, produce, assemble, package, or publish a dataset for training, fine-tuning, evaluation, or benchmarking — including instruction-tuning (SFT), chat/dialogue, preference/DPO/RLHF, pretraining corpora, MCQ benchmarks, QA datasets, or any structured data meant to be loaded by `datasets.load_dataset()` or pushed to HF. Trigger on phrases like "make a dataset from these documents", "build a benchmark", "package this for fine-tuning", "convert this JSONL to a HuggingFace dataset", "produce a DPO/preference/SFT dataset", "I want to publish a dataset". Handles schema design, validation, parquet sharding, statistics, card generation, and provenance — the full production pipeline. Prefer this over hand-rolling dataset code whenever the deliverable is a dataset the user will load, share, or train on.
Produce complete, publish-ready AI/ML datasets in HuggingFace format (parquet shards + README.md with YAML frontmatter + dataset card + provenance script).…
As a process F 50/100 · Will not run — References files that are not bundled: assets/licenses/<spdx-id>.txt
How to improve
- The SKILL.md body is over 5,000 tokens: move reference detail into references/ and load it when needed.
- The text references files that are not there: add them or drop the references.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 0
✓ No critical or high findings
Files scanned: 27. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- warning
body-longSKILL.md body ≈ 6415 tokens (recommended < 5000); move details to references/ - warning
missing-refreference to a missing file: assets/licenses/<spdx-id>.txt
Process rating: all ten parameters 50/100
- 0Tools and files. 1 referenced file(s) missing: assets/licenses/<spdx-id>.txt
- 0Result and completion. Does not say what the result is
- 0Inputs and preconditions. Does not say what the process needs to start
- 70Failures and branches. 6 branches
- 70Execution cost. Instruction body is 6415 tokens
- 85Steps. 105 steps, 1 vague phrases
- 100When it triggers. States when to use and when not to
- 100Consistency. Name and required fields are in place
- 100Running it twice. Mutating operations check current state
- 100Progress reporting. Reports progress
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
- low The response is described with custom markup (13 tags): a typed call is more reliable
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +3Description length 1001: 120–800 characters recommended
- +3Output format is not stated: the model decides each time
- -4Absolute local paths (C:\Users, /home/…): not portable
- +2Single-language instructions
- +5Description quotes 6 example trigger phrases
- +4Description says when NOT to use the skill
- +4Structure: 20 headings
- +3Step-by-step instructions: 105 items
- +4Has examples (8 code blocks)
- +4Reference files are cited in the instructions (7 of 8)
- +3All 5 scripts are documented
- +1License stated
Quality base 70; lint remarks subtract, signals add up to 100. Result: 78.