This skill loads a compact rule set inspired by Martin Kleppmann’s Designing Data-Intensive Applications. It tells a coding agent not to treat writes, queues, caches, replicas and clocks as local, ordered and exactly-once.
It applies to systems where correctness depends on a source of truth, durability, replication, partitioning, event flow, replay or derived data.
What You Get
- A mini rule set with decision rules, trigger rules and a closing checklist for data-system work
- A nano version for tight context budgets and a full version for reference
- A skill description that makes the agent load the rules for consistency, stream and schema tasks
Common Use Cases
- Changing a write path and stating durability, visibility and behavior after a timeout
- Adding a cache, index or read model with defined propagation, lag and rebuild paths
- Making consumers and jobs safe under retry and replay
- Planning a schema or event change across old readers, old writers and in-flight messages
What Makes It Unique
The rules require the agent to state a fault model and the exact guarantee it relies on, such as when a write is durable and when it is visible. Each trigger, from a cache change to a replica read, names the questions to answer before the change goes ahead. It is the one rule set in the repository aimed at distributed data behaviour, with idempotency, ordering scope and schema compatibility treated as first-class concerns.
Under The Hood
The skill is a SKILL.md with a name and description, plus Markdown files holding the mini, nano and full versions of the rules. The SKILL.md tells the agent to read the mini file before design or code decisions and the full file only for harder tradeoffs. It contains no scripts. It does not fetch web pages, call APIs, run commands or write files; the agent only reads the rule text.