Feature Comparison
How rolecraft stacks up against skills (Vercel) and @agentskill.sh/cli (
ags).
At a glance
| rolecraft | skills (Vercel) | @agentskill.sh/cli | |
|---|---|---|---|
| Runtime | Zero-dep Node ESM | 1 dep (Node) | 2 deps (TypeScript) |
| Agent targets | 90 | 72 | 15+ |
| Skill marketplace | — (no marketplace) | agentskill.sh (100K+) | — |
| Publish your own | ❌ | ❌ | ❌ |
| MCP management | ✅ (9 agents) | ❌ | ❌ |
| Bundle / Compose / Diff | ✅ | ❌ | ❌ |
Feature table
| Feature | rolecraft | skills (Vercel) | @agentskill.sh/cli |
|---|---|---|---|
| Zero dependencies | ✅ | ✅ (1 dep) | ❌ (2) |
| Local path install | ✅ 1st class | ✅ | ❌ marketplace only |
| GitHub repo install | ✅ | ✅ | ❌ |
| GitLab / SSH git URL | ✅ | ✅ | ❌ |
| npm package source | ✅ | ✅ | ❌ |
SKILL.md scaffolding (init) | ✅ | ✅ | ❌ |
Skill preview (use) | ✅ | ✅ | ❌ |
Agent auto-detect + install (setup) | ✅ | ❌ | ✅ |
| Skill discovery (search) | ✅ | ✅ | ✅ |
| Interactive TUI search + install | ✅ | ✅ | ❌ |
Bundle install (bundle) | ✅ | ❌ | ✅ (skillset) |
Bundle create (bundle create) | ✅ | ❌ | ❌ |
| Offline capable | ✅ | ✅ | ❌ |
| Project-level install | ✅ | ✅ | ✅ |
| Interactive scope prompt | ✅ | ✅ | ❌ |
Non-interactive flag (--yes/-y) | ✅ | ✅ | ❌ |
Dry-run preview (--dry-run) | ✅ | ❌ | ❌ |
Lockfile collision guard (--frozen-lockfile) | ✅ | ❌ | ❌ |
Content hash verification (verify) | ✅ | ✅ | ❌ |
| Source pinning (commit SHA in lockfile) | ❌ | ❌ | ❌ |
CI-mode re-install (ci) | ✅ | ✅ | ❌ |
Skill update check (check) | ✅ | ❌ | ❌ |
Skill update / re-install (update) | ✅ | ✅ | ❌ |
Symlink install (--symlink) | ✅ | ✅ (default) | ❌ |
Self-upgrade (upgrade) | ✅ | ❌ | ❌ |
| npm provenance | ✅ | ❌ | ❌ |
| Shell completions (bash/zsh/fish) | ✅ | ❌ | ❌ |
In-agent /learn command | ❌ | ❌ | ✅ |
| Skill rating / feedback | ❌ | ❌ | ✅ |
| Skill diff / compose | ✅ | ❌ | ❌ |
System health check (doctor) | ✅ | ❌ | ❌ |
| Watch mode (auto-sync) | ✅ | ❌ | ❌ |
| AGENTS.md XML generation | ✅ | ❌ | ❌ |
| MCP server management | ✅ (9 of 90 agents) | ❌ | ❌ |
| Skill quality test | ✅ | ❌ | ❌ |
Skill conflict detection (doctor --deep) | ✅ | ❌ | ❌ |
| Node.js API | ✅ | ❌ | ❌ |
| Security scanning (0–100) | ✅ | ✅ (Snyk) | ✅ |
| Telemetry / leaderboard | ❌ | ✅ | ❌ |
When to use what
- rolecraft — You need MCP management (for the 9 agents that support it), CI/CD pipelines, skill quality tooling (diff/compose/test). Best for power users and teams.
- skills (Vercel) — You just want the simplest one-liner to install skills. Largest marketplace (90K+ via skills.sh). Best for casual users who trust the Vercel ecosystem.
- @agentskill.sh/cli — You want in-agent discovery via
/learnand skill rating/feedback. Largest marketplace (100K+ via agentskill.sh).
Notable skill collections
These aren't CLI tools — they're curated SKILL.md repositories. rolecraft can install from any of them directly:
| Collection | Skills | Install with rolecraft |
|---|---|---|
| skills.sh | 90,000+ | rolecraft install <owner/repo> |
| agentskill.sh | 100,000+ | rolecraft install <slug> |
| garden-skills (ConardLi) | ~5 curated | rolecraft install ConardLi/garden-skills |
| anthropic/skills | 157K★ repo | rolecraft install anthropic/skills |
Benchmark
rolecraft installs local skills in ~18 ms, against ~4.7 s for npx skills.
That is not like-for-like: the npx figure includes the package download and process startup, and repeated runs varied by roughly a third. Read benchmark/RESULTS.md for the numbers, the method, and what the ratio does and does not measure.
See also
- Benchmark Results — measured install times behind these claims.
- Migrating from Vercel skills — command-by-command mapping if you switch.
- FAQ — short answers to the questions this page raises.