Client Feedback
Clients are not trained to give design feedback and should not have to be. What arrives is a wall of comments in three places, half of them contradictory, some of them symptoms rather than requests — “the header feels cluttered” when the real problem is the type hierarchy two sections down. Translating that into work is a real skill, but the first twenty minutes of it is pure sorting, and that part is tedious every single time.
Skynet does the sorting. It pulls the comments together, strips the duplicates, groups them by the element they touch, and flags the ones that contradict each other or contradict something already approved. What lands on your desk is a structured list. The diagnosis — what the comment actually means and what to do about it — is still your job, which is as it should be.
How it works
Gather every comment into one list
Email replies, file comments, the notes from the review call, the follow-up Slack message. The agent collects them from wherever they landed and deduplicates the ones that are the same note said twice.
Group by what they touch
Comments get clustered by element rather than by who said them. Everything about the nav in one place, everything about the type in another. Patterns show up immediately that were invisible in a chronological thread.
Flag the conflicts and the vague ones
Two comments that cannot both be satisfied get surfaced as a pair. Comments too vague to act on get turned into a specific question you can send back, rather than a guess you make alone.
Turn the clear ones into tasks
Actionable requests become tasks in your tracker, ordered by what you tell the agent matters — deadline, dependency, or client priority. You approve the list before anything is created.
Build it from a prompt
Point it at the mess and tell it what shape you want back.
You get to spend your time deciding what the feedback means instead of assembling it. And the client gets sharper questions back, which usually makes their next round of comments better too.