Install any skill in seconds. Free to start, no credit card required.
Get Started Free →GitHub Projects v2 command center -- create, configure, and manage project boards, views, custom fields, iterations, and item workflows entirely from the editor. Bypasses the drag-and-drop UI that is inaccessible to screen reader users.
.claude/skills/community-access-projects-manager-a952ed/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 47% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 263% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -5% | 0% |
Shared instructions
Skills: github-workflow-standards, github-scanning
You are the Projects Manager. You give screen reader users and keyboard-only users full control over GitHub Projects v2 boards — a feature whose web UI relies heavily on drag-and-drop kanban interactions, visual spatial layouts, and mouse-dependent custom field pickers that are largely inaccessible to assistive technology.
You replace all of that with structured, navigable text output and simple commands.
GitHub Projects v2 boards present severe accessibility barriers:
This agent bypasses all of that by working directly through the GitHub GraphQL API.
Always present project data as structured tables, never as visual boards. For board summaries, use column-grouped lists with clear labels for every item.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-20 | fail→fail | 5,376 | 3,466 | -36% | 1 | 1 | 0% | 957 | 1,150 | +20% | 0 | 0 | — |
case-11 | fail→fail | 5,361 | 12,678 | +136% | 1 | 1 | 0% | 874 | 3,198 | +266% | 0 | 0 | — |
case-12 | pass→pass | 7,308 | 4,918 | -33% | 1 | 1 | 0% | 1,004 | 1,599 | +59% | 0 | 0 | — |
case-18 | pass→pass | 7,641 | 9,402 | +23% | 1 | 1 | 0% | 1,181 | 2,211 | +87% | 0 | 0 | — |
case-01 | fail→fail | 10,814 | 3,606 | -67% | 1 | 1 | 0% | 1,854 | 1,200 | -35% | 0 | 0 | — |
case-02 | pass→fail | 9,639 | 3,905 | -59% | 1 | 1 | 0% | 1,579 | 1,147 | -27% | 0 | 0 | — |
case-03 | fail→pass | 8,656 | 2,942 | -66% | 1 | 1 | 0% | 1,524 | 1,146 | -25% | 0 | 0 | — |
case-04 | fail→pass | 7,941 | 5,839 | -26% | 1 | 1 | 0% | 1,205 | 1,721 | +43% | 0 | 0 | — |
case-05 | fail→pass | 7,605 | 6,123 | -19% | 1 | 1 | 0% | 1,236 | 1,819 | +47% | 0 | 0 | — |
case-06 | pass→pass | 11,003 | 8,063 | -27% | 1 | 1 | 0% | 1,716 | 2,107 | +23% | 0 | 0 | — |
case-19 | fail→fail | 5,166 | 7,828 | +52% | 1 | 1 | 0% | 768 | 1,240 | +61% | 0 | 0 | — |
case-07 | pass→pass | 7,687 | 6,763 | -12% | 1 | 1 | 0% | 1,198 | 1,892 | +58% | 0 | 0 | — |
case-08 | pass→pass | 6,571 | 6,920 | +5% | 1 | 1 | 0% | 1,164 | 2,061 | +77% | 0 | 0 | — |
case-09 | pass→pass | 12,414 | 11,044 | -11% | 1 | 1 | 0% | 2,182 | 2,663 | +22% | 0 | 0 | — |
case-10 | pass→fail | 9,896 | 4,936 | -50% | 1 | 1 | 0% | 1,721 | 1,501 | -13% | 0 | 0 | — |
case-13 | fail→pass | 3,050 | 5,872 | +93% | 1 | 1 | 0% | 460 | 1,672 | +263% | 0 | 0 | — |
case-14 | pass→fail | 8,930 | 5,358 | -40% | 1 | 1 | 0% | 1,550 | 1,349 | -13% | 0 | 0 | — |
case-15 | pass→fail | 8,157 | 3,306 | -59% | 1 | 1 | 0% | 1,412 | 1,103 | -22% | 0 | 0 | — |
case-16 | fail→pass | 8,661 | 4,519 | -48% | 1 | 1 | 0% | 1,605 | 1,519 | -5% | 0 | 0 | — |
case-17 | fail→fail | 6,655 | 4,427 | -33% | 1 | 1 | 0% | 491 | 1,194 | +143% | 0 | 0 | — |
case-21 | pass→fail | 4,932 | 7,469 | +51% | 1 | 1 | 0% | 712 | 1,215 | +71% | 0 | 0 | — |
case-22 | fail→fail | 5,857 | 3,371 | -42% | 1 | 1 | 0% | 982 | 1,123 | +14% | 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 21 counted toward the lift figure. The other 1 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 -8 percentage points is the difference between those two pass rates over the 21 comparable cases. 7 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.