reload-context

name	reload-context
description	Reload full project context at the start of a new conversation by reading living documentation, verifying current state, and identifying next steps.
allowed-tools	Read, Glob, Grep, Task

Reload Project Context

Purpose: Reload full project context at the start of a new conversation by reading living documentation, verifying current implementation state, and identifying next steps.

Instructions

Follow these steps in order to rebuild complete project context:

Phase 1: Read Core Documentation

1. Read project instruction file(s) - CLAUDE.md and/or AGENTS.md

   • If both exist: read both and note any conflicts
   • If only one exists: use that file as the primary instruction source
   • If neither exists: state this explicitly and continue with available project docs
   • Note: Key technologies, development commands, coding patterns
   • Pay special attention to: Documentation standards, styling guidelines, recent improvements

2. List active plans - Find all files in /docs/plans/

   • These are the ONLY plans currently in progress
   • Archived plans are in /docs/archive/ (reference only, not active work)

3. Read each active plan - Understand current objectives and status

   • Note: What's marked as IN PROGRESS
   • Note: What's marked as COMPLETED
   • Note: What's marked as PAUSED
   • Identify: Next immediate steps listed in each plan

Phase 2: Verify Implementation State

4. Use Explore subagent to verify current codebase state
   • Launch with subagent_type: "Explore" and medium thoroughness
   • Ask: "What is the current state of components and features mentioned in active plans?"
   • Cross-reference: Do active plans accurately reflect what's in the codebase?
   • Identify: Any discrepancies between documentation and actual code

Phase 3: Assess Next Steps

5. Use Plan subagent (if needed) to identify optimal next steps
   • Launch with subagent_type: "Plan"
   • Ask: "Based on active plans and current implementation, what are the logical next steps?"
   • Consider: Dependencies, priorities, completion criteria
   • Flag: Any blockers or missing information

Phase 4: Present Comprehensive Summary

Provide a structured summary with these sections:

Project Overview

• Project type and tech stack
• Current development phase
• Key architectural decisions

Active Work

• List each active plan with status
• What's in progress right now
• What's blocked or paused

Recent Completions

• What was recently finished (from "Recent Improvements" in CLAUDE.md and/or AGENTS.md)
• What was archived since last session

Implementation State

• What the Explore subagent found
• Any discrepancies between docs and code
• Current codebase health

Recommended Next Steps

• Prioritized list of next actions
• Dependencies and prerequisites
• Estimated effort/complexity

Flags & Blockers

• Any issues discovered
• Missing information
• Decisions needed

Output Format

Present findings in a clear, scannable format using:

• Headers for organization
• Bullet points for lists
• Status indicators where appropriate
• Code references with file_path:line_number where relevant
• Brief explanations (not just data dumps)

Success Criteria

After running this command, the user should have:

• Complete understanding of current project state
• Clear picture of what's in progress
• Verified implementation matches documentation
• Actionable next steps with context
• Awareness of any blockers or issues

Tips

• Be thorough but concise - Quality over quantity
• Flag discrepancies - If docs don't match code, say so
• Prioritize recent activity - Focus on what's active, not historical
• Ask clarifying questions - If something is unclear after exploration
• Update stale plans - If you find outdated info, note it for cleanup