Qualify composed provider workflows at their actual execution boundary

Before comparing coding agents, exercise the complete staged controller with the actual argument layout, child cwd, sandbox, environment, executable resolution and tool callback. Require actual source reads and tests, valid terminal semantics, source-bound captures and the necessary tool inventory; component tests and a standalone callback can all pass while the composed pipeline delivers no workspace tools. Run advertised baseline commands through that same callback and retain setup failures separately from model accuracy. Use a fresh frozen cohort after infrastructure changes. A toy qualifies the exercised mechanism, not task-solving efficacy or identical vendor harnesses.

Applies when: A controlled software-agent workflow uses this execution or grading boundary. Keywords: coding agent workflow qualification baseline callback MCP Claude Codex