| name | github-issue-complete |
| description | End-to-end automation for the full GitHub issue lifecycle. Use this when the user provides a feature request or bug report and wants to go from zero to a reviewed and refined Pull Request in one autonomous process. |
GitHub Issue Complete Workflow (Full Auto)
This skill provides a "one-stop shop" for the entire software development lifecycle on GitHub. It implements a dual-issue strategy: a Main Issue (human-facing) for project management and a Sub-Issue (agent-facing) for implementation.
📋 Mandatory Task List (task.md)
At the very beginning of Phase 1, the agent MUST initialize and maintain a local task.md file in the workspace containing all steps of the workflow. The agent must update the progress ([ ] to [/] to [x]) continuously and print the updated plan explicitly to the user.
Mandatory task.md Template:
- [ ] Phase 1: Initialization & Strategy
- [ ] Determine/Fetch Main Issue
- [ ] Create/Prepare Human-Facing Main Issue with assignments (read from `package.json` `"agents"` block or fallback `docs/project.json` stakeholders)
- [ ] Create Agent Issue (Sub-Issue, title prefix `[Agent]`) and link natively via helper script or fallback
- [ ] Link Agent Issue inside Main Issue checklist
- [ ] Phase 2: Implementation & Validation
- [ ] Set up Git Worktree and copy `.env` configuration
- [ ] Implement features and refactorings
- [ ] Perform Code Verification (Unit tests, TypeScript compile check, ESLint verification)
- [ ] Perform Local Code Review (utilizing `review-code` skill inside a generalist subagent)
- [ ] Create Pull Request (Title: `feat: ...`, body resolves Agent Issue, references Main Issue)
- [ ] Phase 3: Quality Gate & Review
- [ ] Trigger remote PR review via `review-pull-request` skill
- [ ] Resolve any review findings via `resolve-review` skill
- [ ] Phase 4: Finalization & Handover
- [ ] Update Main Issue checklist to complete
- [ ] Add `test` label to Main Issue
- [ ] Post Browser Click Path and testing instructions comments on Main Issue
- [ ] Ask user to test locally and confirm task completion BEFORE running cleanup
- [ ] Run `npm run log:clear` to clean temporary log files
- [ ] Present final summary with links to Main Issue, Agent Issue, and Pull Request
Phase 1: Initialization & Strategy
-
Analyze User Input: Extract the core objective, any constraints, and relevant context.
-
Initialize Task List: Create task.md with the template above and print the plan to the user.
-
Determine Main Issue:
- Ask the user: "Should I create a new Main Issue for this request, or use an existing one as a reference?"
- YOLO Mode / No Response: If the user is in YOLO mode or provides no specific issue, default to creating a new Main Issue.
- If using an existing issue, fetch its details:
gh issue view <issue-number>.
-
Create/Prepare Main Issue (Human-Facing):
- Goal: This issue tracks the high-level requirement and is used by project managers and testers.
- Content: Must include the user's initial prompt and all relevant information passed to the agent.
- Assignment: Assign the Product Manager and Senior Developer usernames retrieved from
package.json "agents" block or fallback docs/project.json stakeholders (e.g. productManager and seniorDeveloper). If neither is present, default to @me.
- Sub-Task Checklist: Add a section
## Implementation Progress with a checkbox for the sub-issue: - [ ] Agent Issue #<TBD>: [Short Description].
- Action:
- If new:
gh issue create --title "[Feature/Bug] <Title>" --body "<Content>" --assignee "pm-username,dev-username"
- If existing: Update the body to include the new requirement and the checklist.
-
Create and Link Agent Issue (Sub-Issue):
- Goal: This is the technical implementation ticket for the agent, natively linked as a GitHub sub-issue.
- Action:
Phase 2: Implementation & Validation
- Solve the Agent Issue: Activate the
github-issue-solve skill.
- CRITICAL: Pass the Agent Issue Number to
github-issue-solve.
- Follow its workflow for implementation, testing (Vitest, Browser + Logs), and internal review.
- Perform Local Code Review (CRITICAL): Activate the
review-code skill.
- You MUST run the
review-code skill inside a generalist/research subagent to evaluate local changes objectively.
- Meticulously check for compliance with
AGENTS.md, docs/rules/, and look out for critical bugs or code quality issues.
- Any
🔴 Critical or 🟠 High findings must be resolved before proceeding.
- Create Pull Request: The PR should be created as part of the
github-issue-solve workflow.
- Ensure the PR body says
Resolves #<Agent-Issue-Number> and Part of #<Main-Issue-Number>.
Phase 3: Quality Gate & Review
- PR Review: Activate the
review-pull-request skill to review the remote PR.
- Resolution: Activate
resolve-review if changes are requested.
Phase 4: Finalization & Handover
- Update Main Issue:
- Label: Add the
test label to the Main Issue.
- Click Path: Post a comment on the Main Issue with the Browser Click Path for testing (using the PR deployment URL).
- Checklist: Update the checklist in the Main Issue body to mark the sub-issue as complete:
- [x] Agent Issue #<Agent-Issue-Number>.
- Request Local Testing & Confirmation (MANDATORY):
- Stop executing automated steps.
- Explicitly present your final implementation details, screenshots/click-paths, and the URL/port of your local running instance.
- Ask the user to test the changes locally in their browser.
- Seek explicit, written user confirmation that the task is fully complete and that you are authorized to run cleanup.
- Cleanup (Only after confirmation):
- Remove any isolated git worktrees if applicable.
- Run
npm run log:clear to clean temporary log files.
- Final Summary:
- Inform the user that the sub-issue is complete and the Main Issue is ready for testing.
- Provide links to the Main Issue, Agent Issue, and Pull Request.
