boost-skill-cli 1.0.21

boost — Homebrew for AI coding skills

boost — a package manager for AI coding skills

boost is a CLI that finds, installs, and keeps track of AI coding skills pulled from GitHub-hosted registries, then hooks them straight into Claude Code, Windsurf, and Cursor in one pass. Instead of hand-copying skill files into three different agent folders and hoping you remember to update all of them, you run one command.

_{tap a registry → search → install into every agent → doctor — one flow}

What's a "skill," exactly?

A skill is just a SKILL.md file — Markdown plus a bit of YAML up top — that gives an AI agent a repeatable capability: a workflow to follow, a set of house rules, or some domain knowledge it wouldn't otherwise have. Agents like Claude Code pick these up automatically out of ~/.claude/skills/.

Where it gets tedious is everything around that file: tracking down good skills, wiring each one into every agent you use, keeping versions in sync, and handing them off to teammates. That's the part boost takes over — package-manager mechanics for skills, in the same spirit as Homebrew for Mac binaries or npm for JS packages.

Install

Needs Python 3.9+ and git.

pipx install boost-skill-cli        # or: pip install boost-skill-cli
boost --version
boost tap --defaults          # pull in the 5 starter registries

The default install is zero-dependency (pure stdlib). boost search and the boost_search MCP tool rank results with a built-in full-content BM25 engine (run boost reindex once to build the index). For optional dense semantic search, install the [rag] extra and set an embeddings key — retrieval then embeds every skill and ranks by vector similarity, falling back to BM25 whenever the extra or key is absent:

pip install "boost-skill-cli[rag]"            # adds the sqlite-vec vector store
export VOYAGE_API_KEY=...                      # or OPENAI_API_KEY
boost reindex --dense                          # embed chunks into the vector store

Want the whole ecosystem instead of the starter set? boost ships a curated registry catalog — 100+ classified GitHub registries of skills, Cursor/Windsurf rules, and Claude Code workflows (slash commands & subagents), collectively indexing thousands of items. Categories include a curated rag set — official Weaviate, Pinecone, and DSPy skill libraries plus Graph-RAG and agentic-RAG toolkits for building retrieval pipelines:

boost tap --catalog --dry-run                 # browse the classified catalog
boost tap --catalog --type skill --limit 20   # tap the 20 biggest skill packs
boost tap --catalog --type rule               # every rules registry
boost tap --catalog --category rag            # RAG / vector-search skill packs
boost tap --catalog --category security       # filter by category

boost indexes three item kinds out of the same registries: SKILL.md skills, .mdc/.cursorrules/.windsurfrules rules, and command/agent workflows (Markdown under commands/·agents/·workflows/, or carrying subagent frontmatter). Rules and workflows are searchable today; only skills install.

Or run it straight from a checkout — the runtime is stdlib-only, so there's nothing to install beyond the shim:

git clone https://github.com/jonnyeclectic/boost ~/.boost-src
ln -s ~/.boost-src/boost ~/bin/boost        # anywhere on PATH works

Under the hood

boost pulls down registries, indexes them into a local cache, drops installed skills into one canonical store, and then symlinks them out to whichever agents you've got — all of it tracked in a lock file so state stays reproducible.

GitHub registries  ──boost update──▶  ~/.boost/repos/    (shallow clones)
                                     ~/.boost/cache/    (JSON catalogs)
                   ──boost install─▶  ~/.agents/skills/ (canonical store)
                                     .skill-lock.json  (v3 lock file)
                   ────symlinks───▶  ~/.claude/skills/  ~/.windsurf/skills/  ~/.cursor/skills/

The usual workflow

boost search jira          # look across every tapped registry (AI-ranked when available)
boost install my-jira      # copy, link, lock — with a quality score attached
boost doctor               # sanity check: broken links, lock drift, stale taps

# hand the whole setup to a team:
boost bundle dump > Boostfile     # everyone else runs: boost bundle install

75 commands, organized into 8 groups

boost --help prints the full grouped command list; for a visual tour see docs/overview.html.

Group	Commands
Package Management	install · uninstall · sync · update · reinstall · bundle · import · migrate · pin · unpin · snapshot · export
Discovery & Search	search · reindex · discover · recommend · browse · index · trending · stats · count
Skill Information	list · info · cat · edit · preview · explain · log · home · deps · tag
Registry (Taps)	tap · untap · taps · outdated
Intelligence	distill · simulate · infer · absorb · evolve · context · focus · impact
Quality & Health	doctor · lint · audit · verify · drift · test · fingerprint · quarantine · decay · heal · conflict · changelog · attest · health
Configuration	config · clean · create · policy · onboard · completions · schedule · serve · mcp · hooks · bmad · self-update
Team & Collaboration	cohort · profile · protocol · pulse · replay · who

The AI-assisted commands (search --smart, explain, distill, infer, absorb, evolve, simulate, …) call out to the claude CLI when it's available on your PATH (or ANTHROPIC_API_KEY is set), and fall back to plain heuristics when it isn't — so the tool still works without an API key, just less cleverly.

Claude Code hooks & the BMAD Method

boost hooks manages Claude Code hooks in settings.json at either scope — --scope project (./.claude/settings.json) or --scope global (~/.claude/settings.json). boost only ever touches hooks it created (tagged with a # boost:<name> marker in the command), never your own, and snapshots the prior file before each write.

boost hooks add SessionStart -c 'echo hello' -n greet --scope project
boost hooks list
boost hooks remove -n greet --scope project

boost bmad installs and manages the BMAD Method (agentic Analyst / PM / Architect / Dev / QA personas + workflows) as a scope-aware, toggleable startup experience. Provisioning is delegated to the canonical npx bmad-method install (needs Node.js 20.12+); boost owns scope, the startup toggle, and teardown.

boost bmad install --scope project   # skills + per-project _bmad/ runtime
boost bmad install --scope global    # skills into ~/.claude/skills for every session
boost bmad init                      # add the _bmad/ runtime to the current repo
boost bmad startup on                # inject light BMAD orientation on session start
boost bmad startup off               # stop injecting it (skills stay installed)
boost bmad disable / enable          # quarantine / restore skills (recoverable)
boost bmad uninstall                 # delete skills + _bmad/ for a scope
boost bmad doctor                    # what's installed where

When startup is on, a SessionStart hook runs boost bmad orient, which prints a short orientation into the session only while the toggle is enabled. Global installs stage the installer in a temp dir and copy only the bmad-* skills, so $HOME never gets a stray _bmad/ — the workflow runtime stays per-project.

The boost style

The visual identity behind the docs and demos ships as a small, dependency-free design system in style/ — the Aurora living-glass look: a cyan → violet → pink triad on a near-black ground, an ambient aurora that drifts behind the page, and glass cards that light up under the cursor. Drop style/boost.css (and the optional style/boost.js) into any static page; see style/demo.html for the whole system on one page and style/README.md for the tokens and class reference.

Working on boost

# everything runs under a disposable HOME, so nothing touches your real setup:
export HOME=/tmp/boost-sandbox && mkdir -p $HOME
python3 tests/make_fixture.py /tmp/fixture-tap
./boost tap /tmp/fixture-tap
./boost install brainstorming
./boost doctor

Standard-library Python only — no third-party runtime dependencies. Every path under ~/.boost and ~/.agents/skills is resolved from $HOME at call time, which is what makes the sandboxing above possible.

When something misbehaves, boost keeps a rotating diagnostic log at ~/.boost/logs/boost.log and writes a full crash report on any unexpected error. Turn up detail with boost --verbose <cmd> or boost --debug <cmd>, read the trail with boost log --diagnostics, and see docs/DEBUGGING.md for log levels, env vars, crash reports, and the free services that monitor the project.

Test suite

Three layers, all enforced (make check runs the full set; CI runs the same thing):

Layer	What it does	Gate
make test	pytest across tests/unit/ (every core module) and tests/functional/ (drives all 72 commands in-process against sandboxed homes)	≥80% line coverage of boost_cli (fail_under in pyproject.toml)
make smoke	tests/smoke.sh — 152 checks run through the actual ./boost shim (--online also hits real registries)	all pass
make mutation	mutmut mutates boost_cli/core (~2,600 mutants) and reruns the unit suite against each one	≥80% killed (scripts/mutation_gate.py)

Dev setup: make venv (pulls in pytest, coverage, mutmut — the shipped runtime itself stays dependency-free). Every test run uses a throwaway $HOME, so your actual agent configs are never at risk.