harnesslisted
Install: claude install-skill datashaman/code-skills
# Harness
Control surface for the user-scope Claude Code "harness" — feedforward guides (CLAUDE.md, memory), feedback sensors (hooks), and an optional drift-detection loop (snapshot + monthly audit).
The vocabulary follows OpenAI's *Harness engineering* (https://openai.com/index/harness-engineering/) and Martin Fowler's writeup (https://martinfowler.com/articles/harness-engineering.html). Day-to-day patterns are convergent picks from Boris Cherny, Simon Willison, Jesse Vincent (Superpowers), Geoffrey Huntley (Ralph loop), Hamel Husain (eval skills), and Steve Yegge (Gas Town). See README.md for citations.
## Sub-action dispatch
The user invokes this skill, optionally with an action word. Detect intent and run the matching sub-action. If the user says `/harness` without context, run **status** first (it's read-only and informative), then ask which action they want.
| Said by user | Sub-action | Script |
| --------------------------------------------- | ----------- | ------------------------------- |
| "install", "set up", "bootstrap" | `install` | `scripts/install.sh` |
| "uninstall", "remove", "undo" | `uninstall` | `scripts/uninstall.sh` |
| "update", "pull latest templates", "refresh" | `update` | `scripts/update.sh` |
| "doctor", "diagnose", "is it working" | `doctor` | `scripts/doctor.sh` |
| "adopt", "add to existing pr