Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Search GitHub code, repositories, issues, and PRs via MCP
.claude/skills/github-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 76% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 11% | 0% |
bashuv run python -m runtime.harness scripts/mcp/github_search.py \ --type "code" \ --query "your search query"
--type: Search type - code, repos, issues, prs--query: Search query (supports GitHub search syntax)--owner: (optional) Filter by repo owner--repo: (optional) Filter by repo namebash# Search code uv run python -m runtime.harness scripts/mcp/github_search.py \ --type "code" \ --query "authentication language:python" # Search issues uv run python -m runtime.harness scripts/mcp/github_search.py \ --type "issues" \ --query "bug label:critical" \ --owner "anthropics"
Requires github server in mcp_config.json with GITHUB_PERSONAL_ACCESS_TOKEN.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,227 | 5,508 | -12% | 1 | 1 | 0% | 916 | 535 | -42% | 0 | 0 | — |
case-02 | fail→fail | 6,743 | 5,149 | -24% | 1 | 1 | 0% | 1,081 | 589 | -46% | 0 | 0 | — |
case-03 | fail→fail | 8,964 | 5,775 | -36% | 1 | 1 | 0% | 1,668 | 547 | -67% | 0 | 0 | — |
case-04 | pass→fail | 3,282 | 8,019 | +144% | 1 | 1 | 0% | 689 | 1,007 | +46% | 0 | 0 | — |
case-05 | pass→fail | 1,741 | 6,562 | +277% | 1 | 1 | 0% | 198 | 838 | +323% | 0 | 0 | — |
case-06 | pass→fail | 5,948 | 5,596 | -6% | 1 | 1 | 0% | 1,254 | 490 | -61% | 0 | 0 | — |
case-07 | fail→fail | 3,721 | 8,282 | +123% | 1 | 1 | 0% | 753 | 974 | +29% | 0 | 0 | — |
case-08 | fail→fail | 9,896 | 16,507 | +67% | 1 | 1 | 0% | 820 | 1,610 | +96% | 0 | 0 | — |
case-09 | fail→pass | 5,012 | 6,707 | +34% | 1 | 1 | 0% | 804 | 1,419 | +76% | 0 | 0 | — |
case-10 | fail→pass | 4,953 | 1,945 | -61% | 1 | 1 | 0% | 915 | 585 | -36% | 0 | 0 | — |
case-11 | fail→fail | 7,143 | 6,124 | -14% | 1 | 1 | 0% | 1,390 | 703 | -49% | 0 | 0 | — |
case-12 | fail→pass | 8,116 | 9,314 | +15% | 1 | 1 | 0% | 1,614 | 2,174 | +35% | 0 | 0 | — |
case-13 | fail→pass | 5,754 | 2,736 | -52% | 1 | 1 | 0% | 1,090 | 723 | -34% | 0 | 0 | — |
case-14 | fail→pass | 5,528 | 5,630 | +2% | 1 | 1 | 0% | 1,080 | 1,200 | +11% | 0 | 0 | — |
case-15 | fail→pass | 5,049 | 5,202 | +3% | 1 | 1 | 0% | 947 | 1,259 | +33% | 0 | 0 | — |
case-16 | fail→pass | 4,342 | 13,770 | +217% | 1 | 1 | 0% | 809 | 1,911 | +136% | 0 | 0 | — |
case-17 | fail→pass | 4,462 | 8,240 | +85% | 1 | 1 | 0% | 790 | 1,255 | +59% | 0 | 0 | — |
case-18 | fail→pass | 6,729 | 3,918 | -42% | 1 | 1 | 0% | 1,079 | 958 | -11% | 0 | 0 | — |
case-19 | fail→pass | 4,675 | 2,790 | -40% | 1 | 1 | 0% | 972 | 828 | -15% | 0 | 0 | — |
case-20 | fail→fail | 13,589 | 14,912 | +10% | 1 | 1 | 0% | 1,860 | 1,739 | -7% | 0 | 0 | — |
case-21 | fail→fail | 4,289 | 6,256 | +46% | 1 | 1 | 0% | 754 | 705 | -6% | 0 | 0 | — |
case-22 | fail→fail | 2,437 | 8,823 | +262% | 1 | 1 | 0% | 447 | 723 | +62% | 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 12 counted toward the lift figure. The other 10 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 +32 percentage points is the difference between those two pass rates over the 12 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 7/31/2026 | +41% |
Other measured skills in the registry, with their headline benchmark lift.