Agent skill

ocelkit draws its maps from an OCEL 2.0 log, but most systems don't keep one. The events of a process are spread across your code, your database, your logs, and tools like Stripe.

The reconstruction skill helps your coding agent turn them into one log, with you. It's a skill file your agent reads: what a good OCEL 2.0 log looks like, and how to work toward one on any system.

Install the skill

In your project's folder, add the skill with the skills CLI:

$ npx skills add ocelkit/skills --skill building-ocel-logs

The CLI asks which of your agents should get it, such as Claude Code, Codex, or Cursor. Then ask your agent for an OCEL log of the process you want to map.

What your agent does

The skill doesn't do the work for your agent. It's more like a good pair programmer: it knows where the work is going, and how to get there.

• It asks first. A few questions whose answers change the log, such as which part of your system to cover. It finds the objects and events itself.
• It learns your system. It looks at every source, then follows one real case through all of them.
• It proposes a model. The object types, the event types, and what it leaves out and why, before it builds anything.
• It keeps the detail. Lines, items, attributes, and the IDs of your source records stay in the log, so every event traces back to where it came from.
• It writes code. The extraction lives in your repository, and it builds the same log every time it runs.
• It checks its work. It writes checks that could prove the log wrong, such as counting events against the source another way, and runs them.
• It reports honestly. The report starts with the choices most likely to change what you see, and says what it couldn't verify.

What you get

An OCEL 2.0 log saved as SQLite, the code that builds it, and a report. Point ocelkit.config.ts at the log, and you're ready to write your first process file.