How arnab621/typesafe-jev-plugin works
This Claude Code/Cowork plugin supports a two-stage workflow: interactively author reusable TypeSafe Jev solution signatures, then apply a signature to tabular or text inputs and export structured Excel results. The runner’s input, template, API, and output responsibilities are represented at a high level; the excerpt omits parts of its implementation, so uncertain internal call wiring is not shown.
Main entry points: run_dataset.py Dataset runner · create-signature/SKILL.md Create signature · examples Solution signature · run-dataset/SKILL.md Run dataset
Components 14
Signature authoring
- Create signature
skills/create-signature/SKILL.md - Signature designer
agents/signature-designer.md - Solution signature
examples - Signature validator
agents/signature-validator.md
Dataset execution
- Run dataset
skills/run-dataset/SKILL.md - Dataset runner
scripts/run_dataset.py - HTTP API client
scripts/typesafe_client.py - Excel results
Plugin integration
- Signature validation
scripts/validate_signature.py - Save hook
hooks/scripts/validate_on_save.py
Other components
- Plugin user
- CSV, Excel, or text
- TypeSafe Jev API
- Claude plugin host
Connections 12
- Plugin user → Create signature: starts authoring
- Create signature → Signature designer: guides design
- Create signature → Solution signature: creates
- Plugin user → Run dataset: starts run
- Run dataset → Dataset runner: invokes
- CSV, Excel, or text → Dataset runner: provides rows or text
- Solution signature → Dataset runner: configures run
- Dataset runner → TypeSafe Jev API: submits judgments
- Dataset runner → Excel results: writes results
- Signature validator → Signature validation: supports validation
- Claude plugin host → Save hook: dispatches after save
- Save hook → Signature validation: runs validation