
Security News
PHP Foundation Says Chainguard Broke Security Embargo Ahead of Emergency Patches
The PHP Foundation says Chainguard broke a security embargo ahead of critical RCE patches, as leaks disrupted the coordinated release.
@floomhq/starter
Advanced tools
Install curated AI agent skills for Claude Code, Codex, Cursor, OpenCode, and Kimi: zero dependencies, one command.
Install curated AI agent skills for Claude Code, Codex, Cursor, OpenCode, and Kimi: zero dependencies, one command.
npx @floomhq/starter install --global
CLAUDE.md, AGENTS.md, etc.)Starting with 0.2.8, the default selection is all 65 curated skills. The default scope is still project-local: skills are installed under the current directory (./.claude/skills/, ./.codex/skills/, etc.). This keeps your machine clean and lets each project own its own skill set.
To install machine-wide (the old behaviour), pass --global:
# Project-local (default scope): writes all 65 to ./.claude/skills/, ./.codex/skills/, etc.
npx @floomhq/starter install
# Machine-wide: writes all 65 to ~/.claude/skills/, ~/.codex/skills/, etc.
npx @floomhq/starter install --global
The CLI prints which scope it is using on every install.
# Install by profile
npx @floomhq/starter install --profiles core
npx @floomhq/starter install --profiles core,dev,writing
# Install specific skills
npx @floomhq/starter install --skills find-skills,skill-creator
# Install the full curated pack (default)
npx @floomhq/starter install
# Interactive setup (asks your role, picks profiles)
npx @floomhq/starter init
# Show installed skills
npx @floomhq/starter list
# Re-fetch newer skills (preserves your custom edits)
npx @floomhq/starter update
# Remove everything in this scope
npx @floomhq/starter uninstall
# Same thing, explicit form
npx @floomhq/starter remove --all
| Flag | Description |
|---|---|
--profiles <ids> | Comma-separated profile IDs |
--skills <slugs> | Comma-separated skill slugs |
--all | Install all available skills |
--harness <ids> | Target specific agents: claude,codex,cursor,opencode,kimi |
--global | Install machine-wide instead of project-local |
--force | Overwrite existing skills |
--dry-run | Print plan without installing |
--verbose | Print per-skill fetch errors on failure |
--yes | Skip interactive prompts (alias: --non-interactive) |
--version, -v | Print just the version number and exit |
| ID | Description |
|---|---|
core | Skill discovery, task framing, project onboarding |
dev | Code review, tests, security, browser, repo analytics |
writing | Brand voice, email drafts, presentations |
research | Research briefs, citations, enterprise search |
marketing | Landing pages, positioning, customer synthesis |
sales | Outbound, customer context, sales analysis |
ops | Meetings, SOPs, onboarding, file organization |
founder | Strategy, research, customer learning, decisions |
data | Spreadsheets, PDFs, financial QA |
design | Visual QA, browser implementation checks |
video | Transcript, silence, filler-word workflows |
Project-local (default):
| Agent | Skill directory | Activation file |
|---|---|---|
| Claude Code | ./.claude/skills/<slug>/SKILL.md | ./.claude/CLAUDE.md |
| Codex CLI | ./.codex/skills/<slug>/SKILL.md | ./AGENTS.md |
| Cursor | ./.cursor/rules/<slug>.mdc | ./.cursor/rules/floom-skills.mdc |
| OpenCode | ./.opencode/skills/<slug>/SKILL.md | ./.opencode/AGENTS.md |
| Kimi | ./.kimi/skills/<slug>/SKILL.md | ./.kimi/agents/floom-system.md |
Machine-wide (with --global):
| Agent | Skill directory | Activation file |
|---|---|---|
| Claude Code | ~/.claude/skills/<slug>/SKILL.md | ~/.claude/CLAUDE.md |
| Codex CLI | ~/.codex/skills/<slug>/SKILL.md | ~/.codex/AGENTS.md |
| Cursor | ~/.cursor/rules/<slug>.mdc | ~/.cursor/rules/floom-skills.mdc |
| OpenCode | ~/.config/opencode/skills/<slug>/SKILL.md | ~/.config/opencode/AGENTS.md |
| Kimi | ~/.kimi/skills/<slug>/SKILL.md | ~/.kimi/agents/floom-system.md |
npx @floomhq/starter update
Re-fetches the manifest and per-skill content, re-installs only the skills whose remote version is newer than what you have locally. Files you have customised are NOT overwritten (the collision check from 0.2.1 preserves your versions).
For a fresh install with the latest CLI plus the latest manifest:
npx @floomhq/starter@latest install
# Remove everything in the current scope
npx @floomhq/starter uninstall
# Same thing, explicit form
npx @floomhq/starter remove --all
# Remove specific skills
npx @floomhq/starter remove --skills find-skills,skill-creator
# Remove an entire profile
npx @floomhq/starter remove --profiles dev
uninstall and remove delete the SKILL.md files the installer wrote, clear the Floom activation block from CLAUDE.md / AGENTS.md, and delete the local manifest. Files the collision check skipped (your custom versions) are not touched. Both commands respect the --global flag, so npx @floomhq/starter uninstall --global cleans up a previous machine-wide install.
npx @floomhq/starter list
Lists every skill installed via Floom in the current scope, which profile it belongs to, and which agent paths it was written to. Pass --global to list machine-wide installs.
Ask your agent: "find a skill for this task": the find-skills skill searches only the Floom-installed skills in this project or machine scope. It does not call external registries unless you explicitly ask for broader discovery.
Or: "write a plan for this work": the workplan skill guides the planning flow.
Claude Code, Codex, Cursor, Kimi, OpenCode. Gemini is not supported (the CLI will print a clear error and exit 1 if you pass --harness gemini).
@floomhq/starter uses only Node.js built-ins (Node 18+). No install overhead.
floom.dev: the skill registry for AI coding agents.
FAQs
Install curated AI agent skills for Claude Code, Codex, Cursor, OpenCode, and Kimi: zero dependencies, one command.
We found that @floomhq/starter demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.

Security News
The PHP Foundation says Chainguard broke a security embargo ahead of critical RCE patches, as leaks disrupted the coordinated release.

Security News
PHP releases emergency patches for two critical RCE vulnerabilities as agencies urge organizations to take unpatched sites offline.

Threat Research
/Security News
A supply chain attack compromised LiteLLM 1.107.0 on PyPI and Docker, spreading malware that steals cloud credentials, developer secrets, and AI provider keys.