Developer tools plugin with skills for journaling, productivity, and autonomous development workflows.
/plugin install ac-dev-tools@alteredcraft-pluginsA development journaling skill that helps developers reflect on and document their work using Claude Code activity data.
Usage:
Skill(ac-dev-tools:journal)
When triggered, the skill will:
- Ask for a time period (if not specified)
- Gather activity data from Claude Code directories
- Synthesize a narrative journal entry
- Offer to refine until you're satisfied
Features:
- User-centric narratives: Captures what you accomplished, not AI actions
- Selective highlighting: Gauges significance to highlight meaningful work
- Multi-day summaries: Provides highlights and day-by-day breakdowns
- Iterative refinement: Always offers to revise before finalizing
Data Sources:
~/.claude/history.jsonl- Activity timeline~/.claude/projects/- Session transcripts~/.claude/file-history/- File modifications~/.claude/todos/- Task completion status~/.claude/stats-cache.json- Usage metrics
See skills/journal/references/claude-code-directory.md for detailed documentation of these data structures.
Example Output:
## Dev Journal: January 3, 2026
You had a focused day on the Claude Explorer API. The highlight was implementing the activity summary endpoint—something you'd been planning for a while.
**Claude Explorer** - Implemented the activity summary endpoint
- Designed the response schema for cross-project aggregation
- Added date range filtering with proper timezone handling
- ✅ Add /activity/summary endpoint
- ✅ Add /activity endpoint for detailed timeline
*Also accomplished: Fixed a null pointer in the todos parser*Sets up autonomous coding tasks using the Ralph Wiggum methodology — structured requirements gathering (Phase 1) and implementation planning (Phase 2), then hands off to a building loop.
Usage:
Skill(ac-dev-tools:ralph-method) [task-name-and-description]
When triggered, the skill will:
- Check if you're on
main— offer to create a feature branch - Interview you about Jobs to Be Done
- Decompose into topics (one-sentence test)
- Generate specs in
specs/[task-name]/ - Create detailed implementation plan
- Deploy
BUILD_PROMPT.mdandralph.shto project root - Hand off with instructions to run Phase 3
Output Structure:
your-project/
├── BUILD_PROMPT.md # Instructions for each loop iteration
├── ralph.sh # Loop runner script
└── specs/
└── [task-name]/ # Namespaced task specs
├── PRD.md # Product requirements with checkboxes
├── IMPLEMENTATION_PLAN.md
├── BACKPRESSURE.md # Feedback loops (tests, lint, build)
├── AGENTS.md # Task-specific learnings
└── 01-topic.md # Topic specifications
Running Phase 3:
./ralph.sh # Lists tasks, prompts for selection
./ralph.sh [task-name] # Run specific task directly
./ralph.sh [task-name] 20 # With custom max iterationsSee skills/ralph-method/README.md for detailed documentation.
MIT