@pennyfarthing/core 7.7.0 → 7.8.1

package/pennyfarthing-dist/commands/git-cleanup.md

@@ -1,340 +1,75 @@
 ---
 description: Clean up git repos by organizing changes into proper commits/branches by initiative
+workflow: git-cleanup
 ---
 
 # Git Cleanup Command
 
 Analyze and organize uncommitted changes across all repos into proper commits and branches based on initiative/feature groupings.
 
-## Git Workflow Rules
-
-**CRITICAL: Never commit directly to develop. Branch protection hooks will reject direct commits.**
-
-All changes MUST follow this workflow:
-1. Create a branch from develop
-2. Commit changes to the branch
-3. Merge to develop locally
-4. Push develop (branches don't need to be pushed)
-
-## Analysis Phase
-
-### 1. Gather Current State and Check for Unpushed Commits
-
-```bash
-./scripts/run.sh git/git-status-all.sh
-```
-
-This shows branch, changes, and unpushed commits for all repos.
-
-### 2. Check Recent Commits for Patterns
-
-```bash
-echo "=== Recent Commit Patterns ==="
-git log --oneline -10 | head -10
-echo ""
-echo "=== Recent Branches ==="
-git branch --sort=-committerdate | head -10
-```
-
-### 3. Check Active Work Sessions
-
-```bash
-echo "=== Active Work ==="
-if [ -f ".session/{STORY_ID}-session.md" ]; then
-  head -30 .session/{STORY_ID}-session.md
-else
-  echo "No active work session"
-fi
-```
-
-### 4. Check Worktree Status
-
-```bash
-echo "=== Worktree Status ==="
-./scripts/run.sh git/worktree-manager.sh status
-```
-
-This shows all active worktrees with their branches and uncommitted changes.
-
-### 5. Check Epics for Story Context
-
-```bash
-echo "=== Active Epics ==="
-head -60 docs/epics.md
-```
-
-## Categorization Guidelines
-
-Group changes by these initiative types (based on conventional commits):
+## BikeLane Workflow
 
-| Prefix | Type | Branch Pattern | Example |
-|--------|------|----------------|---------|
-| `docs:` | Documentation | `docs/description` | docs/update-epics |
-| `chore:` | Maintenance | `chore/description` | chore/sprint-cleanup |
-| `chore(sprint):` | Sprint tracking | `chore/sprint-update` | - |
-| `chore(pennyfarthing):` | Pennyfarthing config | `chore/pennyfarthing-cleanup` | - |
-| `feat:` | New feature | `feat/story-id-desc` | feat/4-12-risk-scoring |
-| `fix:` | Bug fix | `fix/issue-desc` | fix/validation-error |
-| `refactor:` | Code improvement | `refactor/description` | refactor/seed-data |
+This command uses the **git-cleanup** stepped workflow:
 
-## Common Change Categories
+| Step | Name | Purpose |
+|------|------|---------|
+| 1 | Analyze | Gather git status across all repos |
+| 2 | Categorize | Group changes by initiative type |
+| 3 | Execute | Create branches, commit, merge |
+| 4 | Verify | Confirm clean state, optionally push |
+| 5 | Complete | Summary and next steps |
 
-### Sprint/Docs Changes
-Files: `docs/*.md`, `sprint/*.yaml`, `sprint/*.md`
-- Branch: `chore/sprint-cleanup` or `docs/update-[topic]`
-- Commit: `chore(sprint): update sprint tracking` or `docs: update documentation`
+## Quick Start
 
-### Pennyfarthing Configuration
-Files: `.claude/**/*`
-- Branch: `chore/pennyfarthing-[description]`
-- Commit: `chore(pennyfarthing): description`
-
-### Feature Work
-Files: `internal/**`, `src/**`, `migrations/**`
-- Branch: `feat/[story-id]-description`
-- Commit: `feat: description`
-
-### Seed/Test Data
-Files: `internal/seed/**`, `cmd/seed/**`
-- Branch: `chore/seed-improvements` or part of feature branch
-- Commit: `chore(seed): description`
-
-### Bug Fixes
-Files: Various
-- Branch: `fix/[issue-description]`
-- Commit: `fix: description`
-
-## Execution Phase - Branch Workflow
-
-**For each group of changes, follow this exact workflow:**
-
-### Step 1: Stash All Changes First
-```bash
-# Stash everything to start clean
-git stash push -m "cleanup-wip"
-```
-
-### Step 2: Create Branch from Develop
-```bash
-# Ensure on develop and up to date
-git checkout develop
-git pull origin develop
-
-# Create feature branch
-git checkout -b type/description
-```
-
-### Step 3: Apply Relevant Changes
-```bash
-# Pop stash
-git stash pop
-
-# Stage only files for this group
-git add <specific-files>
-
-# Stash remaining changes for next group
-git stash push -m "remaining-cleanup"
-```
-
-### Step 4: Commit
-```bash
-# Commit with proper message
-git commit -m "$(cat <<'EOF'
-type(scope): description
-
-Details if needed.
-
-🤖 Generated with [Claude Code](https://claude.com/claude-code)
-
-Co-Authored-By: Claude <noreply@anthropic.com>
-EOF
-)"
-```
-
-### Step 5: Merge to Develop
-```bash
-# Switch to develop
-git checkout develop
+Run `/git-cleanup` to start the workflow.
 
-# Merge the branch
-git merge type/description
-
-# Delete local branch
-git branch -d type/description
-```
-
-### Step 6: Repeat for Next Group
-```bash
-# Pop remaining stash and repeat from Step 2
-git stash pop
-```
-
-## Multi-Repo Cleanup Pattern
-
-When cleaning up changes across multiple repos (configured in `.claude/project/repos.yaml`):
-
-```bash
-# Source repo utilities
-source $CLAUDE_PROJECT_DIR/scripts/repo-utils.sh
-
-# For each configured repo, run the same workflow
-for_each_repo '
-  # ... branch workflow ...
-  git status --short
-'
-```
-
-Or manually iterate:
-
-```bash
-source $CLAUDE_PROJECT_DIR/scripts/repo-utils.sh
-for repo in $(get_repos); do
-  repo_path=$(get_repo_full_path "$repo")
-  echo "=== Processing $repo ==="
-  cd "$repo_path"
-  # ... branch workflow ...
-  cd -
-done
-```
-
-## Interactive Cleanup Process
-
-When running this command, Claude should:
+## Git Workflow Rules
 
-1. **Show current state** - Display all uncommitted changes grouped by repo
-2. **Check for unpushed commits** - Identify any commits on develop not yet pushed
-3. **Analyze changes** - Look for patterns in file paths, recent commits, and epics
-4. **Propose groupings** - Suggest how to organize changes:
-   ```
-   Group 1: Sprint Updates
-   - docs/epics.md
-   - sprint/current-sprint.yaml
-   → Branch: chore/sprint-cleanup
-   → Commit: "chore(sprint): clean up sprint tracking files"
+**CRITICAL: Never commit directly to develop.** Branch protection hooks will reject direct commits.
 
-   Group 2: Migration Fix
-   - migrations/056_*.sql
-   → Branch: fix/risk-scoring-migration
-   → Commit: "fix(db): add missing risk scoring migration"
+All changes follow this pattern:
+1. Create branch from develop
+2. Commit changes to branch
+3. Merge to develop locally
+4. Push develop (branches stay local)
 
-   Group 3: Seed Improvements
-   - internal/seed/*.go
-   → Branch: chore/seed-improvements
-   → Commit: "chore(seed): improve demo data generation"
-   ```
-4. **Ask for confirmation** - Let user approve or modify groupings
-5. **Execute** - Create branches, commit, merge to develop for each group
-6. **Push develop** - Push all merged changes at the end
-7. **Verify** - Show clean git status at the end
+## Categorization Reference
 
-## Safety Rules
+| Prefix | Type | Branch Pattern |
+|--------|------|----------------|
+| `docs:` | Documentation | `docs/description` |
+| `chore:` | Maintenance | `chore/description` |
+| `chore(sprint):` | Sprint tracking | `chore/sprint-update` |
+| `feat:` | New feature | `feat/story-id-desc` |
+| `fix:` | Bug fix | `fix/issue-desc` |
+| `refactor:` | Code improvement | `refactor/description` |
+| `test:` | Test changes | `test/description` |
 
-- **NEVER commit directly to develop** - hooks will reject it
-- **Never force push**
-- **Never commit secrets** (.env, credentials, etc.)
-- **Show diff before committing**
-- **Respect .gitignore**
-- **Handle unpushed commits on develop before starting cleanup**
+## Manual Quick Reference
 
-## Quick Reference
+For quick cleanup without the full workflow:
 
 ```bash
-# View what needs cleanup (all repos)
+# View what needs cleanup
 ./scripts/run.sh git/git-status-all.sh
 
-# Create cleanup branch
+# Standard cleanup sequence
+git stash push -m "cleanup-wip"
 git checkout develop && git pull
 git checkout -b chore/cleanup-$(date +%Y%m%d)
-
-# Stage specific files
-git add "docs/*.md"
-git add "sprint/*.yaml"
-
-# Commit with proper message (use heredoc for multiline)
-git commit -m "$(cat <<'EOF'
-chore(sprint): update sprint status
-
-🤖 Generated with [Claude Code](https://claude.com/claude-code)
-
-Co-Authored-By: Claude <noreply@anthropic.com>
-EOF
-)"
-
-# Merge to develop (local)
+git stash pop
+git add <files>
+git commit -m "chore: description"
 git checkout develop
 git merge chore/cleanup-$(date +%Y%m%d)
 git branch -d chore/cleanup-$(date +%Y%m%d)
-
-# Push develop after all merges complete
 git push origin develop
 ```
 
-## Branch Cleanup
-
-Stale feature branches can accumulate over time. Include branch cleanup as part of git hygiene.
-
-### Check for Stale Branches
-
-```bash
-echo "=== Branch Status ==="
-
-./scripts/run.sh misc/check-status.sh
-```
-
-### Remove Merged Branches
-
-**Only remove branches that are fully merged:**
-
-```bash
-# Source repo utilities
-source $CLAUDE_PROJECT_DIR/scripts/repo-utils.sh
-
-# For each configured repo
-for repo in $(get_repos); do
-  repo_path=$(get_repo_full_path "$repo")
-  echo "=== $repo ==="
-  cd "$repo_path"
-
-  git checkout develop
-  git pull origin develop
-
-  # List branches merged into develop
-  git branch --merged develop | grep -v "develop\|main"
-
-  # Delete merged branches
-  git branch --merged develop | grep -v "develop\|main" | xargs -r git branch -d
-
-  cd -
-done
-```
-
-### Branch Cleanup Criteria
-
-| Branch Type | When to Remove |
-|-------------|----------------|
-| `feat/*` | After PR merged to develop |
-| `fix/*` | After PR merged to develop |
-| `chore/*` | After PR merged to develop |
-
-**Note:** Never delete `main` or `develop` branches.
-
-## Post-Cleanup Verification
-
-```bash
-echo "=== Final State ==="
-./scripts/run.sh git/git-status-all.sh
-echo ""
-echo "=== Branches ==="
-
-# Source repo utilities
-source $CLAUDE_PROJECT_DIR/scripts/repo-utils.sh
-
-for repo in $(get_repos); do
-  repo_path=$(get_repo_full_path "$repo")
-  echo "$repo:" && git -C "$repo_path" branch
-done
-```
-
----
+## Safety Rules
 
-**Start by analyzing, then propose groupings, then execute with user approval. Always use branches!**
+- **NEVER** commit directly to develop
+- **NEVER** force push
+- **NEVER** commit secrets (.env, credentials)
+- **ALWAYS** show diff before committing
+- **ALWAYS** use branches

package/pennyfarthing-dist/commands/solo.md

@@ -249,6 +249,37 @@ It goes directly to a file, then jq reads it safely.
 
 ## Step 6: Invoke Judge Skill
 
+**Detect SWE-bench scenarios for deterministic evaluation:**
+
+Check if the scenario is from SWE-bench by looking at its path or category:
+```python
+is_swebench = (
+    'swe-bench' in scenario_path.lower() or
+    scenario.get('category') == 'swe-bench' or
+    scenario.get('source') == 'swe-bench'
+)
+```
+
+**If SWE-bench scenario:**
+
+Use deterministic Python-based evaluation instead of LLM-as-judge:
+
+```bash
+# Save response to temp file for Python judge
+echo '{"result": "{RESPONSE}"}' > /tmp/solo_response_$$.json
+
+# Run SWE-bench judge (deterministic scoring against ground truth)
+python3 .pennyfarthing/scripts/test/swebench-judge.py {scenario_name} /tmp/solo_response_$$.json
+```
+
+The Python script returns:
+- `total`: Score out of 100
+- `scores`: Breakdown by category (root_cause, fix_quality, completeness, persona)
+- `details`: Specific findings and matches
+
+**If standard scenario (non-SWE-bench):**
+
+Use LLM-as-judge:
 ```
 /judge --mode solo --data {
   "spec": "{spec}",

package/pennyfarthing-dist/guides/patterns/approval-gates-pattern.md

@@ -757,7 +757,7 @@ When implementing approval gates:
 - `Task` with `subagent_type`: Delegated gate execution
 
 ### Reflector Hook
-- `question-reflector-check.mjs`: Enforces CYCLIST marker before AskUserQuestion
+- `question_reflector_check.py`: Enforces CYCLIST marker before AskUserQuestion
 - Markers: `<!-- CYCLIST:CHOICES:... -->`, `<!-- CYCLIST:QUESTION:yesno|open -->`
 
 ---