From uberclaw
Inventory the user's Claude Code skills, agents, and commands into a Markdown profile UberClaw consumes. Use when the user runs /uberclaw:train to refresh the per-project profile at <project>/.claude/uberclaw/profile.md.
How this skill is triggered — by the user, by Claude, or both
Slash command
/uberclaw:trainThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Refresh the UberClaw training profile from the user's Claude Code
Refresh the UberClaw training profile from the user's Claude Code
installation. Reads only frontmatter from each *.md / SKILL.md —
never body content — so it is always safe to run.
bin/uberclaw-train
Writes to <project>/.claude/uberclaw/profile.md. The file is Markdown,
user-editable, and intended to be committed alongside the rest of
.claude/. Each project gets its own profile so a train run in project
A does not stomp on project B.
bin/uberclaw-train --global
Writes to ${HOME}/.claude/uberclaw/profile.md for skills/agents/commands the
user wants surfaced across every project.
The profile has two clearly-marked regions:
A <!-- BEGIN GENERATED -->…<!-- END GENERATED --> block. train
regenerates this block on every invocation.
A free-form ## User notes section. train preserves this section
byte-for-byte across runs. Hand-edit it freely:
priority: speed, skip the architecture pass."After invocation, this skill renders the helper's stdout summary back to the user — the resolved output path plus per-scope counts.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub humanassisted/uberclaw