A collection of reusable Claude Code skills and slash commands, packaged as an installable plugin. Use it to back up your custom workflows, sync them across machines, and share them.
A prompt-pack build pipeline — run them in order — plus a few utilities.
| Command | What it does |
|---|---|
/create-prompt-packs [spec] |
Decompose specs (PRD, roadmap, agents.md) into milestone-by-milestone build plans that /build can execute. Run this first. |
/verify-coverage [scope] |
Audit spec→pack ownership: every spec'd capability has an owning pack or an explicit deferred home. The per-pack commands can't see features that landed in no pack — this can. Run after generation, after spec changes, and at phase boundaries. |
/build [pack-id] |
Execute a prompt pack — build all features in a milestone with tests, committing incrementally. Resume-safe. |
/verify-build [pack-id] |
Audit build completeness: every deliverable, test, and acceptance criterion specified in the pack. Run before /verify-wiring. |
/verify-wiring [pack-id] |
Audit integration wiring: every feature is reachable end-to-end (navigation, API callers, schedulers, UI), not just present. |
/review-externally [scope] |
Run an external code review (via Codex) on prompt-pack commits or recent changes. |
/build-verify-review [flags] |
Orchestrator: run build → verify-build → verify-wiring → review-externally for every pack, fixing gaps autonomously. |
/self-improvement-loop |
Add a self-improvement loop to a project's CLAUDE.md — learning behaviors, decision logging, commit/changelog practices. Idempotent. |
/changelog [description] |
Update CHANGELOG.md (Keep a Changelog format) from your recent changes. |
/smart-commit [message] |
Stage changes, update the changelog, and create a well-formatted commit. |
| Skill | Triggers when |
|---|---|
learn |
Something surprising happens, an approach fails, a useful pattern emerges, or you point out something to remember — captures it to memory / decisions / CLAUDE.md. |
manage-skill |
A reusable workflow pattern emerges, or you ask to create/update/list/delete a skill. |
Skill vs. command: a command lives in
commands/<name>.md, is invoked by you as/<name>, and supports$ARGUMENTS. A skill lives inskills/<name>/SKILL.md, is auto-invoked by Claude based on itsdescription, and does not take$ARGUMENTS.
/plugin marketplace add salian/claude-toolkit
/plugin install salian-claude-toolkit@salian-claude-toolkit
Once installed, the commands and skills are available in every project on that machine. Re-run /plugin marketplace update salian-claude-toolkit to pull the latest.
Copy or symlink the pieces into your Claude Code config:
git clone https://github.com/salian/claude-toolkit.git
cd claude-toolkit
# user-level (all projects)
ln -s "$PWD/commands"/*.md ~/.claude/commands/
ln -s "$PWD/skills"/* ~/.claude/skills/Or drop them under a single project's .claude/commands/ and .claude/skills/ instead.
These are Claude Code–specific in format (SKILL.md, slash commands with $ARGUMENTS, allowed-tools frontmatter) — they won't run as-is in other AI coding tools. Some commands also assume a few conventions: a memory/ directory with MEMORY.md, docs/decisions/ for ADRs, and a prompt-pack build flow. Where those are absent the skills adapt or ask; nothing breaks.
.claude-plugin/
plugin.json # plugin manifest
marketplace.json # marketplace listing
commands/ # slash commands (build pipeline + utilities)
skills/ # auto-invoked skills (learn, manage-skill)