Skip to content

Glossary

Epic. One feature request on a GitHub issue that Forge plans, builds, tests and delivers.

Story. One slice of an epic with its own tests and branch.

Brownfield. An existing codebase, as opposed to a new project.

Characterization test (pinning test). A test that records what the code does today, bugs included, so a change in behavior is caught.

Parity. The new code gives the same results as the old code on the same tests. See Parity Proof.

Slice-by-slice (strangler). Replacing a system in pieces while it keeps working, instead of rewriting it all at once.

Mechanical, assisted, human-required. The three classes of migration difficulty. See The Assessment.

Windows runner. The Windows container where .NET Framework code is built and tested.

BYOK. Bring your own key for the AI vendor. See Bring your own AI key.

Evidence pack. A bundle of change, review, test, cost and deploy records with an integrity hash. See Security overview.

Guardrail (Sign). A project rule every agent reads and every reviewer enforces.

Repo skill. A markdown file of team conventions under .forge/skills/ that is injected into every agent.

Drift probe. A scheduled check against an external source of truth, such as a .NET target framework reaching end of support. It files one issue per distinct finding.

MCP. Model Context Protocol, a standard way for an agent to call tools such as CloudWatch.

Early access. Built and tested, enabled per project, not yet proven at scale.