One pipeline primitive, many languages
Reusable CI should standardize security and delivery behavior across languages while leaving language-specific build logic explicit.
Standardize policy, not every command
Go, Node, Python, Rust, and .NET projects do not need identical build steps. They do need consistent expectations around checkout, dependency integrity, test evidence, artifact naming, vulnerability scanning, provenance, and publishing. The reusable primitive should encode those shared controls without pretending the ecosystems are the same.
A thin language adapter can own toolchain setup and native build conventions. A shared orchestration layer can own permissions, security gates, artifact flow, and deployment contracts. That split makes the standard useful without turning it into a growing switch statement.
Make the contract narrow
Inputs should describe intent—runtime version, artifact location, publish mode—not inject arbitrary shell. Outputs should be stable enough for later stages to consume without knowing how a language produced them. Permissions should default to read-only and expand only in the job that signs or publishes.
Pinning actions, separating untrusted pull-request work from privileged release work, and passing immutable artifacts between stages do more for supply-chain quality than adding another scanner with broad credentials.
Adoption is part of the design
A reusable workflow succeeds when repositories can adopt it incrementally and failures remain diagnosable. Version the contract, publish migration notes, expose the underlying tool output, and keep escape hatches explicit. Centralization should reduce duplicated security decisions, not hide build behavior from the teams who own it.