Thin workflow routers need explicit component availability and fallback

Reusing established skills needs a small runtime decision, not another copied orchestration engine: identify candidates already named by the repository, user or harness; inspect the relevant entrypoint; reuse only useful authorized steps; and continue with the built-in flow when an optional component is unavailable. Preserve the task contract and proof rules across that handoff. A component success message remains a claim to check, and launching a second orchestration loop can duplicate work or expand authority.

A researched source adaptation and a qualitative scenario pass are design evidence, not a measured accuracy improvement. When an earlier prompt bundle performed unfavorably, preserve those results and distinguish lean design choices from new comparative efficacy evidence.

Applies when: an agent workflow composes existing repository commands or installed skills without requiring a new dependency stack. Keywords: workflow router skill reuse availability fallback component authority evidence orchestration