An extended health check for Claude Code. Runs the built-in /doctor, then audits your workspace against the 6 then & now context-engineering shifts Anthropic shipped with the Claude 5 models.
Anthropic cut ~80% of the Claude Code system prompt for their newest models. Most workspaces are still written for the old ones. This skill finds where yours is.
git clone https://github.com/robonuggets/doctor-plusCopy the skill folder into your project (or your user-level skills):
cp -r doctor-plus/.claude/skills/doctor-plus <your-project>/.claude/skills/doctor-plusThen in Claude Code:
/doctor-plus
Part 1 - the standard checkup. Runs the built-in claude doctor: install health, dead weight, CLAUDE.md trimming, slow hooks.
Part 2 - the 6 then & now shifts. Audits your CLAUDE.md, skills, rules, and memory against the principles Anthropic now prompts by:
| # | Then | Now |
|---|---|---|
| 1 | Give Claude rules | Let Claude use judgement |
| 2 | Give Claude examples | Design interfaces |
| 3 | Put it all upfront | Use progressive disclosure |
| 4 | Repeat yourself | Simple tool descriptions |
| 5 | Memory in CLAUDE.md | Auto-memory |
| 6 | Simple specs | Rich references |
It audits the principles, not keyword patterns - and it knows safety rules (approvals, backups, destructive actions, money) are hard on purpose and never suggests softening them.
Part 3 - report first. One findings table with the file and passage behind every flag. Nothing changes until you approve.
doctor-plus/
├── .claude/skills/doctor-plus/SKILL.md # the skill (the entry point)
├── assets/ # banner + infographic
├── README.md
└── LICENSE # CC BY 4.0
/doctor-plus- "run a then and now audit on this workspace"
- "context checkup - report only, don't fix anything"
The 6 shifts come from The new rules of context engineering for Claude 5 models by Thariq (@trq212) at Anthropic.
Created by Jay from RoboLabs. Learn more at RoboNuggets.
CC BY 4.0 - free to use with attribution.