aiblueprint-cli 1.3.0 → 1.3.2

package/claude-code-config/commands/tasks/run-task.md

@@ -0,0 +1,220 @@
+---
+allowed-tools: Bash(gh :*), Bash(git :*)
+argument-hint: <issue-number|issue-url|file-path>
+description: Execute GitHub issues or task files with full EPCT workflow and PR creation
+---
+
+<objective>
+Execute #$ARGUMENTS using adaptive APEX workflow (Analyze → Plan → Execute → eXamine) with automatic PR creation.
+
+This command handles any task input: GitHub issues, file paths, or direct descriptions. It adapts complexity based on task size - lightweight for bug fixes, comprehensive for features. Optimized for CI/automated environments.
+</objective>
+
+<context>
+Current branch: !`git branch --show-current`
+Git status: !`git status --short`
+Recent commits: !`git log --oneline -5 2>/dev/null || echo "No commits yet"`
+</context>
+
+<process>
+
+## 0. INITIALIZE
+
+**Parse input and setup environment:**
+
+1. **Detect input type** from #$ARGUMENTS:
+   - **Issue number** (e.g., `123`): Fetch with `gh issue view 123`
+   - **Issue URL** (e.g., `https://github.com/.../issues/123`): Extract number, fetch with `gh issue view`
+   - **File path** (e.g., `tasks/feature.md`): Read file content
+   - **Direct description**: Use as-is
+
+2. **For GitHub issues**:
+   - Add processing label: `gh issue edit <number> --add-label "in-progress"`
+   - Extract: title, description, labels, comments
+
+3. **Assess task complexity** (determines workflow depth):
+   - **Simple** (typo, config change, small fix): Skip to Execute phase
+   - **Medium** (bug fix, small feature): Light exploration + execute
+   - **Complex** (new feature, refactor): Full APEX workflow
+
+4. **Setup branch**:
+   - If on `main`/`master`: Create branch `git checkout -b feat/<task-slug>`
+   - If on feature branch with uncommitted changes: Ask to continue or stash
+   - If on feature branch with commits: Ask to continue or create new branch
+
+## 1. ANALYZE (skip for simple tasks)
+
+**Goal**: Gather context efficiently
+
+- Launch **parallel** `explore-codebase` agents for:
+  - Related files and patterns
+  - Similar implementations to use as examples
+  - Test file locations
+- Use `explore-docs` agent if external library knowledge needed
+- Use `websearch` agent only if truly necessary
+
+**Output**: Mental map of files to modify and patterns to follow
+
+## 2. PLAN (skip for simple tasks)
+
+**Goal**: Define implementation strategy
+
+1. Create concise implementation plan:
+   - Files to modify/create
+   - Key changes per file
+   - Test approach
+
+2. **For GitHub issues**: Post plan as comment
+
+   ```
+   gh issue comment <number> --body "## Implementation Plan
+
+   - [ ] Change 1
+   - [ ] Change 2
+   - [ ] Tests
+
+   Starting implementation..."
+   ```
+
+3. **If ambiguity exists**: Ask user ONE focused question, then proceed
+
+## 3. EXECUTE
+
+**Goal**: Implement changes following codebase patterns
+
+1. **Create TodoWrite** with specific tasks (1 per file change)
+
+2. **For each change**:
+   - Read target file first
+   - Match existing code style exactly
+   - Use clear names over comments
+   - Stay strictly in scope
+
+3. **Code quality rules**:
+   - NO comments unless truly necessary
+   - NO refactoring beyond task scope
+   - NO extra features
+   - Run formatters after changes
+
+## 4. EXAMINE
+
+**Goal**: Validate implementation
+
+1. **Check package.json** for available scripts
+
+2. **Run validation** (in order, stop on failure):
+
+   ```bash
+   npm run format 2>/dev/null || true
+   npm run lint
+   npm run typecheck
+   ```
+
+3. **Run relevant tests only**:
+   - Find test file for modified code
+   - Run: `npm test -- --testPathPattern="<pattern>"` or equivalent
+   - Don't run full suite unless necessary
+
+4. **If validation fails**:
+   - Fix errors immediately
+   - Re-run validation
+   - If stuck after 2 attempts: report blocker to user
+
+## 5. FINALIZE (always execute)
+
+**Goal**: Create PR and update tracking
+
+1. **Stage and commit**:
+
+   ```bash
+   git add -A
+   git commit -m "<type>(<scope>): <description>
+
+   <body if needed>
+
+   Closes #<issue-number>"
+   ```
+
+   Commit types: `feat`, `fix`, `refactor`, `docs`, `test`, `chore`
+
+2. **Push and create PR**:
+
+   ```bash
+   git push -u origin HEAD
+   gh pr create --fill --body "## Summary
+   <what was done>
+
+   ## Changes
+   - <change 1>
+   - <change 2>
+
+   ## Test Plan
+   - [x] Lint passes
+   - [x] Types pass
+   - [x] Relevant tests pass
+
+   Closes #<issue-number>"
+   ```
+
+3. **Update GitHub issue** (if applicable):
+
+   ```bash
+   gh issue comment <number> --body "✅ Implementation complete
+
+   **PR**: <pr-url>
+   **Changes**: <brief summary>
+
+   Ready for review."
+   ```
+
+4. **Return PR URL** to user
+
+</process>
+
+<adaptive_behavior>
+**Task size detection heuristics:**
+
+| Signal             | Likely Simple                   | Likely Complex                         |
+| ------------------ | ------------------------------- | -------------------------------------- |
+| Keywords           | "typo", "rename", "update text" | "implement", "add feature", "refactor" |
+| Scope              | Single file mentioned           | Multiple components                    |
+| Issue labels       | `bug`, `docs`, `chore`          | `feature`, `enhancement`               |
+| Description length | < 100 chars                     | > 500 chars                            |
+
+**Workflow adaptation:**
+
+- Simple: Initialize → Execute → Examine → Finalize
+- Medium: Initialize → Light Analyze → Execute → Examine → Finalize
+- Complex: Full workflow with all phases
+  </adaptive_behavior>
+
+<error_handling>
+**Common blockers and responses:**
+
+- **Lint/type errors**: Fix immediately, max 3 attempts
+- **Test failures**: Debug, fix, document if unfixable
+- **Merge conflicts**: Report to user, don't auto-resolve
+- **Missing context**: Use explore agents, ask user if still unclear
+- **CI blocked**: Wait and retry once, then report
+
+**If blocked**: Report clearly what's blocking and what's needed
+</error_handling>
+
+<success_criteria>
+
+- Task requirements fully addressed
+- Code passes lint, typecheck, relevant tests
+- PR created with clear description
+- GitHub issue updated (if applicable)
+- PR URL returned to user
+  </success_criteria>
+
+<execution_rules>
+
+- **ULTRA THINK** at phase transitions
+- **Parallel execution** for speed (agents, independent commands)
+- **Stay in scope** - implement exactly what's requested
+- **Always create PR** - even for small fixes
+- **Priority**: Correctness > Completeness > Speed
+- **Automated-friendly**: Minimize user prompts, proceed with sensible defaults
+  </execution_rules>

package/claude-code-config/commands/{watch-ci.md → utils/watch-ci.md}

@@ -41,3 +41,7 @@ You are a CI monitoring specialist. Watch pipelines and fix failures automatical
 ## Priority
 
 Fix accuracy > Speed > Minimal commits. Ensure CI is truly green.
+
+---
+
+User: #$ARGUMENTS