Repo Curriculum Planner

Generates structured learning/improvement plans based on repo state and session brain.

Two Usage Modes

Mode 1: Propose in Chat (default)

When a user asks for planning help, propose tasks directly in the conversation without writing files. This is the default behavior for quick suggestions.

Examples:

• "what should I work on next?" → Suggest 2-3 high-priority tasks in chat
• "what's the roadmap?" → Describe the plan conversationally
• "help me prioritize" → Discuss tradeoffs and recommend order

Mode 2: Write Artifacts to Disk

Only write to .claude/voyager/ when the user explicitly asks to persist the plan:

• "write the curriculum to disk"
• "save this plan"
• "generate curriculum files"
• "materialize the plan"

This writes:

• .claude/voyager/curriculum.json — structured task backlog
• .claude/voyager/curriculum.md — human-readable plan

Trigger Phrases

Use this skill when you hear:

• "what should I work on next?"
• "create a plan"
• "what's the roadmap?"
• "help me onboard"
• "plan the work"
• "prioritize features"
• "what tasks remain?"
• "make an onboarding plan"
• "roadmap this refactor"

On-Disk Plan Artifacts

When persisted, the plan lives at:

• .claude/voyager/curriculum.json — structured JSON with full task details
• .claude/voyager/curriculum.md — rendered Markdown for human review

To refresh an existing plan, run:

voyager curriculum plan

To preview without saving:

voyager curriculum plan --dry-run

Task Structure

Each task includes:

• id — unique identifier (e.g., t01)
• title — short action-oriented title
• why — rationale for the task
• estimated_scope — trivial, small, medium, or large
• acceptance_criteria — how to know it's done
• suggested_files — where to look
• commands_to_run — verification commands
• depends_on — prerequisite task IDs
• status — pending, in_progress, done, or blocked

Updating Task Status

Edit curriculum.json directly to update task status:

• Change "status": "pending" to "in_progress" when starting
• Change to "done" when complete

See reference.md for implementation details and schema reference.