Fast local search across agent skill registries with quality filtering - only returns skills with score >= 80 by default.
Based on analysis of 4,784 skills from 5 registries: Analyzing 4,784 AI Agent Skills →
- Quality filtering: Filters out low-quality skills (score < 80) by default
--min-scoreflag: Override the minimum quality threshold (use 0 to show all)- Fast local search: Tantivy (BM25) full-text search engine
- Git-based sync: Clones repos locally for instant access
- Multiple registries: Searches clawdhub, anthropic, openai, and jo skills
- Trust indicators:
[✓]for trusted,[⚠]for untrusted - Quality scores: Shows
[Q:score]for each skill
- git: Required for cloning skill registries
Download from releases:
| Platform | Binary |
|---|---|
| macOS (Apple Silicon) | safe-skill-search-aarch64-apple-darwin.tar.gz |
| macOS (Intel) | safe-skill-search-x86_64-apple-darwin.tar.gz |
| Linux (x86_64) | safe-skill-search-x86_64-unknown-linux-gnu.tar.gz |
| Linux (ARM64) | safe-skill-search-aarch64-unknown-linux-gnu.tar.gz |
| Windows (x86_64) | safe-skill-search-x86_64-pc-windows-msvc.zip |
git clone https://github.com/jo-inc/safe-skill-search
cd safe-skill-search
cargo install --path .# Search for skills (only shows quality score >= 80)
safe-skill-search search "browser automation"
# Show all skills regardless of quality score
safe-skill-search search "browser automation" --min-score 0
# Show only premium quality skills (score >= 90)
safe-skill-search search "browser automation" --min-score 90
# Search with JSON output
safe-skill-search search "pdf" --json
# Filter by registry
safe-skill-search search "pdf" --registry anthropic
# Only trusted skills (anthropic + openai official)
safe-skill-search search "document" --trusted
# Show top skills by stars (quality filtered)
safe-skill-search top
# Show skill details including quality score
safe-skill-search show trello
# Get install URL for a skill
safe-skill-search url trello
# Force resync from GitHub
safe-skill-search sync --forceQuality scores are based on the skills-db analysis which evaluated 4,784 skills from 5 registries.
Scoring criteria:
- Documentation completeness
- Code examples
- Workflow structure
- Error handling
- Best practices
Scores range from 0-100, with 80+ being considered high-quality.
| Registry | Source | Skills | Trust |
|---|---|---|---|
| clawdhub | github.com/openclaw/skills | ~3400 | ⚠ Community |
| anthropic | github.com/anthropics/skills | ~16 | ✓ Official |
| openai | github.com/openai/skills/.curated | ~31 | ✓ Official |
| openai-experimental | github.com/openai/skills/.experimental | varies | ⚠ Experimental |
| jo | github.com/jo-inc/skills | varies | ✓ Official |
All data stored in ~/.local/share/skill-search/:
skills.db- SQLite database with skill metadataindex/- Tantivy full-text search indexrepos/- Cloned git repositories (~100MB total)
cargo build --release
# Binary at target/release/safe-skill-searchMIT