The skill in practice
Effective Engineering
Design the system. Make its contracts hold.
Effective Engineering works on the software itself: architecture, APIs, data behavior, TypeScript, Rust, and the tests that protect them. It starts from the repository and the actual failure consequences, then makes the technical choice explicit enough for the next maintainer to understand.
The range
What it covers
- Architecture and APIs
- System boundaries, quality attributes, operability, request and event contracts, compatibility, retries, and recovery.
- Data systems
- Ownership, invariants, transactions, consistency, replication, schema evolution, backfills, cutovers, and reconciliation.
- TypeScript depth
- Types, public module APIs, async behavior, typed errors, boundary validation, and compiler configuration.
- Rust depth
- Ownership, lifetimes, crate structure, public APIs, domain types, errors, async boundaries, and concurrency.
- Rust performance and safety
- Profiling, allocation, memory layout, binary size, unsafe code, FFI, SIMD, atomics, and written safety arguments.
- Tests and benchmarks
- Focused regression tests, invariants, failure paths, CLI contracts, flake diagnosis, and measurements that report their environment and variance.
From request to result
How it works
Locate the contract
Read the implementation and identify the invariant, boundary, or failure consequence that matters.
Choose and implement
Compare the viable options and make the smallest change that satisfies the actual constraints.
Produce evidence
Use focused tests or measurements that can distinguish the claimed improvement from the old behavior.
The result: Code and system boundaries with explicit behavior, failure contracts, and evidence.
Ask in your own words
Try a real task.
Give the agent the relevant files, source material, and constraints. Start with an outcome like one of these:
Should this stay a modular monolith or split into services? Argue from our actual load, ownership, and failure consequences, not from a pattern.
Review this crate's public API before 1.0. I care about lifetimes leaking internals and about panic paths I have not documented.
This `as unknown as T` cast is holding a boundary together. Replace it with a narrowing that actually proves the type, or tell me why it cannot be done.
Choose the right discipline
When to use it
This is the discipline for non-frontend architecture and code. Effective Web owns browser-facing TypeScript and frontend testing. Effective Delivery runs established repository checks, maintains PRs, updates dependencies, and writes technical documentation. Benchmark guidance does not claim to cover load, soak, or stress-test execution.
For the complete workflow and routing rules, read the agent instructions on GitHub .
Add this discipline
Install Effective Engineering
With Node.js installed, run this command for an agent that supports Agent Skills:
npx skills add sebastian-software/effective-agent --skill effective-engineering
Then describe your task in the agent. This skill works on its own; add another when the work crosses disciplines.
For managed installations, pinned selections, and multiple agent targets, see all installation options or the DALO setup guide.