Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create GitHub Pull Request from specification using pull_request_template.md. Use when: spec needs to be converted to PR, spec is ready for review/merge, need to automate PR creation from specification file with template-based body and title.
.claude/skills/giuseppe-trisciuoglio-create-pr-from-spec/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 64% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 93% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 355% | 0% |
Create a GitHub Pull Request for a specification using the pull_request_template.md template located at ${workspaceFolder}/.github/pull_request_template.md.
This skill automates the creation of GitHub Pull Requests directly from specifications. It follows a structured process:
Key Benefits:
Use this skill when:
Trigger phrases:
Before using this skill, ensure:
main, develop).github/pull_request_template.md exists in the repository1. Analyze specification template
${workspaceFolder}/.github/pull_request_template.md2. Create pull request draft
create_pull_request tool to create draft PR to target branchget_pull_request to prevent duplicates3. Get pull request changes
get_pull_request_diff tool to analyze differences4. Update pull request
update_pull_request tool to populate PR body and title5. Mark ready for review
update_pull_request tool to change state from draft to ready for review6. Assign pull request
get_me to retrieve current user informationupdate_issue tool to assign PR to creator7. Return pull request URL
bashInput: Create PR from spec to main branch Process: 1. Analyze pull_request_template.md 2. Create draft PR to main 3. Check for existing PRs 4. Update PR title: "feat: [Feature Name from Spec]" 5. Update PR body with template sections 6. Mark as ready for review 7. Assign to creator Output: https://github.com/user/repo/pull/123
markdownPR Title: feat: Implement user authentication system PR Body: ## Description Implements JWT-based authentication with token refresh mechanism ## Related Issue Closes #456 ## Changes - Added JWT middleware - Implemented token validation - Added refresh token endpoint ## Testing - Unit tests for auth middleware - Integration tests for token endpoints ## Checklist - [x] Tests pass - [x] Documentation updated - [x] No breaking changes
bashInput: Create PR from spec Check: PR for current branch already exists? → Yes: Report error, don't create duplicate → No: Proceed with PR creation Output: "PR already exists at https://github.com/user/repo/pull/789"
⚠️ Important:
.github/pull_request_template.md🚫 Limitations:
search - Analyze specification template requirementscreate_pull_request - Create new PR in draft modeget_pull_request - Check for existing PRs before creationget_pull_request_diff - Analyze PR changesupdate_pull_request - Update PR title, body, and stateupdate_issue - Assign PR to creatorget_me - Retrieve current user information| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 4,531 | 5,015 | +11% | 1 | 1 | 0% | 504 | 1,851 | +267% | 0 | 0 | — |
case-02 | fail→fail | 5,900 | 5,316 | -10% | 1 | 1 | 0% | 933 | 1,716 | +84% | 0 | 0 | — |
case-03 | fail→fail | 6,946 | 5,527 | -20% | 1 | 1 | 0% | 1,232 | 1,857 | +51% | 0 | 0 | — |
case-04 | pass→pass | 9,287 | 4,158 | -55% | 1 | 1 | 0% | 1,551 | 2,137 | +38% | 0 | 0 | — |
case-05 | pass→pass | 7,682 | 1,959 | -74% | 1 | 1 | 0% | 1,339 | 1,708 | +28% | 0 | 0 | — |
case-11 | pass→pass | 10,899 | 1,351 | -88% | 1 | 1 | 0% | 935 | 1,610 | +72% | 0 | 0 | — |
case-06 | fail→pass | 11,760 | 3,172 | -73% | 1 | 1 | 0% | 1,775 | 1,932 | +9% | 0 | 0 | — |
case-07 | fail→pass | 6,075 | 2,244 | -63% | 1 | 1 | 0% | 1,095 | 1,800 | +64% | 0 | 0 | — |
case-08 | fail→pass | 5,803 | 2,012 | -65% | 1 | 1 | 0% | 901 | 1,742 | +93% | 0 | 0 | — |
case-09 | fail→fail | 13,559 | 4,906 | -64% | 1 | 1 | 0% | 2,278 | 2,177 | -4% | 0 | 0 | — |
case-10 | pass→pass | 9,762 | 1,943 | -80% | 1 | 1 | 0% | 1,476 | 1,690 | +14% | 0 | 0 | — |
case-12 | pass→pass | 6,656 | 1,990 | -70% | 1 | 1 | 0% | 911 | 1,676 | +84% | 0 | 0 | — |
case-13 | fail→fail | 4,366 | 2,599 | -40% | 1 | 1 | 0% | 604 | 1,858 | +208% | 0 | 0 | — |
case-14 | fail→pass | 11,003 | 2,380 | -78% | 1 | 1 | 0% | 1,173 | 1,772 | +51% | 0 | 0 | — |
case-15 | fail→pass | 2,916 | 3,820 | +31% | 1 | 1 | 0% | 450 | 2,048 | +355% | 0 | 0 | — |
case-16 | pass→pass | 10,532 | 3,318 | -68% | 1 | 1 | 0% | 1,704 | 1,966 | +15% | 0 | 0 | — |
case-17 | fail→pass | 10,637 | 2,288 | -78% | 1 | 1 | 0% | 1,641 | 1,825 | +11% | 0 | 0 | — |
case-18 | fail→pass | 11,542 | 2,218 | -81% | 1 | 1 | 0% | 1,950 | 1,789 | -8% | 0 | 0 | — |
case-19 | pass→pass | 9,998 | 2,338 | -77% | 1 | 1 | 0% | 1,416 | 1,734 | +22% | 0 | 0 | — |
case-20 | pass→fail | 8,070 | 41,408 | +413% | 1 | 1 | 0% | 813 | 2,024 | +149% | 0 | 0 | — |
case-21 | pass→pass | 5,018 | 4,403 | -12% | 1 | 1 | 0% | 778 | 2,214 | +185% | 0 | 0 | — |
case-22 | pass→pass | 6,732 | 10,078 | +50% | 1 | 1 | 0% | 1,126 | 2,466 | +119% | 0 | 0 | — |
case-23 | fail→fail | 12,120 | 5,124 | -58% | 1 | 1 | 0% | 2,395 | 1,637 | -32% | 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. 23 cases were attempted, and 18 counted toward the lift figure. The other 5 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 +26 percentage points is the difference between those two pass rates over the 18 comparable cases. 1 case got worse with the skill loaded, and it is 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.