tomevault-io Search Skill
Discover existing skills from curated sources, evaluate quality with toothbrush filter, output comparison table. Philosophy: inspect, adapt, own — never import blindly 🪥
Workflow
- Parse query from
$ARGUMENTS(e.g., "pdf converter", "git workflow") - Search tiered sources (parallel where possible):
| Tier | Source | Method |
|---|---|---|
| 1 | Anthropic official repos | WebSearch: "$QUERY claude code skill site:github.com/anthropics" |
| 1 | VoltAgent curated | WebSearch: "$QUERY site:github.com/VoltAgent/awesome-agent-skills" |
| 2 | GitHub community | WebSearch: "$QUERY claude code SKILL.md site:github.com" |
| 2 | Awesome lists | WebSearch: "$QUERY awesome-claude-skills OR awesome-claude-code" |
| 3 | SkillHub rated | WebFetch: skillhub.club search |
| 3 | Broad web | WebSearch: "$QUERY claude code skill 2026" |
- Fetch SKILL.md for top candidates via
WebFetchon raw GitHub URLs - Apply toothbrush filter — see
reference.mdfor rubric details:- Token efficiency: <500 ideal, hard fail >2000
- Clear triggers: keywords + file types required
- Single responsibility: one capability only
- Anthropic patterns + progressive disclosure (soft)
- Output comparison report:
🔍 Found N skills for "$QUERY":
| Skill | Source | ~Tokens | Quality | Verdict |
|-------|--------|---------|---------|---------|
| name | source | ~N | ⭐⭐⭐⭐ | 🟢/🟡/🔴 Study/Adapt/Ignore |
💡 Recommended approach:
- [What to study from best matches]
- [What patterns to adapt]
- [What to ignore and why]
📎 Key patterns found:
- [Reusable patterns across results]
Rules
- Output verdicts (study/adapt/ignore), NOT install commands
- Skip tiers 2-3 if tier 1 has strong matches
- Fetch actual SKILL.md source when possible — evaluate code not marketing
- See
reference.mdfor full source registry and quality rubric weights
Converted and distributed by TomeVault — claim your Tome and manage your conversions.