Webflow Marketplace / 07 of 08
Turn review experience into reusable preparation
I turned recurring Marketplace app review problems into shareable skills and checklists developers can use before submission and after receiving findings.
Selected professional work by Micah Johnson for Webflow, presented as individual portfolio work. Webflow did not commission this work through CREATE SOMETHING. Individual contributions are described within a wider team and platform context.
Contribution focus: Guidance design, workflow routing, evidence checklists, packaging, and consistency evaluation
Original source and actual static evaluation output. No live model behavior is demonstrated.
Silent recording · Select captions for visual guidance. Open the MP4 ↗
The problem
Skills
A general coding assistant can produce plausible advice that misses the requirements of a particular platform. Developers also need different guidance when they are preparing a first submission and when they already have specific findings to resolve.
The contribution
The work behind the interface.
The preparation skill walks through building and checking a Marketplace app. The remediation skill starts from issued findings, preserves their wording and acceptance criteria, routes fixes to the owning layer, and builds evidence for resubmission readiness. The two workflows deliberately route different requests.
The packages include checklists, references, evaluation cases, and static consistency harnesses. They separate published platform requirements from additional security or engineering guidance. Supporting reviewer skills describe queue operation, analysis interpretation, write guardrails, and fallback when a review flow cannot proceed.
The key decision
What needed to stay clear.
Treat the guidance as a maintained product. Keep its claims grounded, preserve the finding contract, and distinguish source checks, artifact checks, runtime observation, submission, and reviewer acceptance.
The documented result
What the evidence shows.
The new recording replays original skill excerpts and actual local static-harness output: 158 preparation checks and 97 remediation checks passed. These results establish grounding and package consistency, not model behavior or review success.
The source and static-evaluation replay contains no live model conversation or behavioral evaluation. Historical package download verification remains a separate earlier result.
Portfolio evidence captured October 2026. Historical implementation and recordings do not establish the current state of every production system.