Dataset and input factory
Skill alainlebret/claude-agents/higher-ed-teaching-agents/skills/dataset-and-input-factory
Modular skill and workflow ecosystems for AI coding assistants. Structured, multi-agent pipelines for specialized professional domains.
npx -y skills add alainlebret/claude-agents --skill dataset-and-input-factoryAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Generates realistic synthetic datasets, input files, traces, logs, images, CSV files, graphs, and edge-case inputs for computing labs, projects, and exams. Use when pedagogical artifacts need representative inputs or benchmark data.
SKILL.md
2.9 KB, as published. Nobody here has run it
Dataset and Input Factory
Goal
Produce synthetic, realistic, and pedagogically appropriate input data for a computing assignment, TP, exam, or benchmark. Data must be varied enough to test student code thoroughly, and must never expose real personal or sensitive information.
Inputs
mission.json— level, domain, constraints, evaluation_modestatement.md— the assignment subject; data must be consistent with what is describedrubric.md— grading criteria; each data category should exercise at least one criterion
Output contract
Place all generated data under data/ or tests/cases/ as appropriate.
| File | Description |
|---|---|
data/README.md | Describes each dataset: format, generation method, known properties, intended test coverage |
data/<name>.<ext> | One or more synthetic input files (CSV, JSON, log, binary, etc.) |
data/generate.py (or .sh) | Reproducible generation script so data can be regenerated with a fixed seed |
data/edge-cases/ | Subdirectory with adversarial or boundary inputs |
Dataset categories to produce
For each assignment, generate at minimum:
- Nominal case — realistic, well-formed input that exercises the happy path
- Large case — input at or near the stated size limit to test performance
- Empty / minimal case — empty file, zero elements, or minimum valid input
- Malformed case — syntactically invalid input to test error handling
- Edge case — boundary values (max int, unicode characters, trailing newlines, binary data mixed with text)
Domain-specific guidance
| Domain | Typical data formats | Notes |
|---|---|---|
| Systems / C | Log files, named pipes, binary streams | Use ASCII-safe content; avoid locale-specific characters unless stated |
| Python data | CSV, JSON, plain text | Use random.seed(42) for reproducibility |
| SQL | SQL dump or CSV for import | Include NULL values and duplicate keys |
| Networking | PCAP excerpts or HTTP request logs | Use RFC-example addresses (192.0.2.x) |
| Embedded / real-time | Sensor traces, CAN frames | Include timing jitter and out-of-order packets |
Rules
- All data must be synthetic — never use real user data, real logs, or scraped datasets.
- The generation script must be deterministic given a fixed seed.
- Document the statistical properties of each dataset (size, value range, class distribution).
- Data must be consistent with the domain scenario described in
statement.md. - Do not generate data that would require students to install additional tools to process (e.g. proprietary binary formats).