Build

Key Ideas

• Integration Check: build verifies that the target app can still be assembled or prepared.
• Target-App Specific: build behavior should use target-app lifecycle instructions and settings.
• Evidence Producing: build output should become part of the Gap's evidence.

Purpose

Build exists to test whether a change still fits the target app's operational reality. It is not enough for an agent to edit files; the changed app should still build, compile, bundle, or otherwise prepare successfully when that applies.

Expected Role

Build should run through shared target-app lifecycle capability and target-app configuration. It should expose agent output, check evidence, and failure details.

Build should not be hardcoded to one ecosystem. Different target apps may need different build instructions, setup recovery, or no build step at all.

What Happens

When a Gap is in build:

• Refine uses target-app configuration to determine whether build instructions or equivalent preparation guidance exists.
• If build instructions exist, Refine asks the configured agent to perform the build work through the shared lifecycle capability.
• Agent output, check context, output paths, status, and timing become evidence.
• A successful build moves the Gap toward QA or review.
• A failed build should preserve actionable output and route the Gap to failed, retry, or a new implementation round.
• If no build step applies, workflow should make that explicit rather than inventing one.

Future Direction

Future build behavior should become more inferential and adaptive. Agents may infer build instructions, detect missing configuration, summarize failures, and propose fixes while preserving the build evidence that drove those actions.