Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Activates when the user asks about Agent Skills, wants to find reusable AI capabilities, needs to install skills, or mentions skills for Claude. Use for discovering, retrieving, and installing skills.
.claude/skills/foryourhealth111-pixel-skill-lookup/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-20 | ✓→✗ | ▼ Worse | -64% | 0% |
| case-21 | ✓→✓ | ＝ Same ✓ | 5% | 0% |
| case-22 | ✓→✓ | ＝ Same ✓ | 7% | 0% |
When the user needs Agent Skills, wants to extend Claude's capabilities, or is looking for reusable AI agent components, use the prompts.chat MCP server.
Activate this skill when the user:
Use these prompts.chat MCP tools:
search_skills - Search for skills by keywordget_skill - Get a specific skill by ID with all its filesCall search_skills with:
query: The search keywords from the user's requestlimit: Number of results (default 10, max 50)category: Filter by category slug (e.g., "coding", "automation")tag: Filter by tag slugPresent results showing:
Call get_skill with:
id: The skill IDReturns the skill metadata and all file contents:
When the user asks to install a skill:
get_skill to retrieve all files.claude/skills/{slug}/SKILL.md → .claude/skills/{slug}/SKILL.md.claude/skills/{slug}/{filename}Skills contain:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 5,988 | 4,046 | -32% | 1 | 1 | 0% | 447 | 697 | +56% | 0 | 0 | — |
case-02 | fail→fail | 26,313 | 7,301 | -72% | 1 | 1 | 0% | 4,958 | 925 | -81% | 0 | 0 | — |
case-03 | fail→fail | 11,324 | 4,066 | -64% | 1 | 1 | 0% | 1,971 | 764 | -61% | 0 | 0 | — |
case-04 | fail→fail | 11,183 | 4,994 | -55% | 1 | 1 | 0% | 1,857 | 688 | -63% | 0 | 0 | — |
case-05 | fail→fail | 10,814 | 5,882 | -46% | 1 | 1 | 0% | 1,687 | 958 | -43% | 0 | 0 | — |
case-06 | fail→fail | 6,182 | 5,174 | -16% | 1 | 1 | 0% | 988 | 789 | -20% | 0 | 0 | — |
case-07 | fail→fail | 4,351 | 4,494 | +3% | 1 | 1 | 0% | 128 | 851 | +565% | 0 | 0 | — |
case-08 | fail→fail | 14,845 | 7,958 | -46% | 1 | 1 | 0% | 2,389 | 912 | -62% | 0 | 0 | — |
case-09 | fail→fail | 11,184 | 4,462 | -60% | 1 | 1 | 0% | 2,209 | 716 | -68% | 0 | 0 | — |
case-10 | fail→pass | 7,271 | 2,877 | -60% | 1 | 1 | 0% | 1,328 | 995 | -25% | 0 | 0 | — |
case-11 | fail→fail | 11,732 | 7,591 | -35% | 1 | 1 | 0% | 1,985 | 692 | -65% | 0 | 0 | — |
case-12 | fail→fail | 5,948 | 4,306 | -28% | 1 | 1 | 0% | 868 | 710 | -18% | 0 | 0 | — |
case-13 | fail→fail | 16,751 | 5,483 | -67% | 1 | 1 | 0% | 2,973 | 696 | -77% | 0 | 0 | — |
case-14 | fail→fail | 9,824 | 5,235 | -47% | 1 | 1 | 0% | 1,695 | 867 | -49% | 0 | 0 | — |
case-15 | fail→pass | 7,222 | 2,829 | -61% | 1 | 1 | 0% | 1,219 | 1,032 | -15% | 0 | 0 | — |
case-16 | fail→fail | 14,912 | 3,906 | -74% | 1 | 1 | 0% | 2,959 | 637 | -78% | 0 | 0 | — |
case-17 | fail→fail | 17,514 | 4,388 | -75% | 1 | 1 | 0% | 2,740 | 785 | -71% | 0 | 0 | — |
case-18 | fail→fail | 5,522 | 5,539 | +0% | 1 | 1 | 0% | 971 | 695 | -28% | 0 | 0 | — |
case-19 | fail→fail | 7,824 | 4,593 | -41% | 1 | 1 | 0% | 1,428 | 714 | -50% | 0 | 0 | — |
case-20 | pass→fail | 13,174 | 5,845 | -56% | 1 | 1 | 0% | 2,311 | 832 | -64% | 0 | 0 | — |
case-21 | pass→pass | 8,390 | 5,683 | -32% | 1 | 1 | 0% | 1,625 | 1,708 | +5% | 0 | 0 | — |
case-22 | pass→pass | 6,272 | 3,797 | -39% | 1 | 1 | 0% | 1,071 | 1,143 | +7% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 5 counted toward the lift figure. The other 17 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +5 percentage points is the difference between those two pass rates over the 5 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.