Ben Honda
Posts and shiplogs tagged with “claude”.
Anthropic's new design tool ships with anti-slop guardrails. They don't fire on their own. The DESIGN.md, system prompt, named aesthetic, and reference-mix that actually do.
Skills are markdown files Claude reads when relevant. The description field decides whether one ever triggers. What changed in Opus 4.7, and the trap that silently disables half your skills.
Both tools default to the same generic AI aesthetic: Inter, purple-on-white, three-up icon cards. A practitioner playbook for steering each off it, and the cross-tool file doing most of the work.
A working playbook for stopping Claude from defaulting to LinkedIn-thinkpiece prose: the leverage points, the banned-words block, the three-pass workflow, and the before/after.