post-compact-context-reload

Roger-Parkinson-EHP's avatarfrom Roger-Parkinson-EHP

Full context reload. Use after /clear, after compaction, or at session start. Loads conversation history, indexes docs, maps code, scans configs, checks git. Gets you immediately helpful.

⭐0stars🔀0forks📁View on GitHub🕐Updated Jan 2, 2026

When & Why to Use This Skill

This Claude skill automates the full restoration of project context, enabling the AI to quickly regain a comprehensive understanding of conversation history, codebase structure, documentation, and git state. It is specifically designed to solve the 'cold start' problem after context clearing, compaction, or session resets, ensuring the agent remains immediately productive by systematically indexing the local environment.

Use Cases

  • Resuming complex development tasks after a context compaction or '/clear' command to free up token space while maintaining project awareness.
  • Initializing a new coding session by automatically mapping the codebase, identifying core abstractions, and reading project-specific rules (CLAUDE.md).
  • Synchronizing the AI's state with the current Git branch and uncommitted changes to ensure accurate assistance with version control and ongoing work.
  • Rapidly scanning configuration files (like pyproject.toml or package.json) and documentation to understand the build system and architectural constraints of a new repository.
namepost-compact-context-reload
descriptionFull context reload. Use after /clear, after compaction, or at session start. Loads conversation history, indexes docs, maps code, scans configs, checks git. Gets you immediately helpful.
allowed-toolsBash, Read, Grep, Glob

Context Reload

The master command for getting up to speed.

Use after:

  • /clear command
  • Context compaction
  • Session start
  • Anytime context feels thin

The Reload Sequence

Execute in order. Each builds on the previous.


1. CONVERSATION HISTORY (Who was I? What was I doing?)

cd ~/.claude/services
python3 conversation-search.py --list-sessions | head -5

Then search for the current project:

python3 conversation-search.py "PROJECT_NAME" --max 3 --full

Capture: Last task, pending items, user preferences.


2. DOCUMENTATION INDEX (What should I know?)

find . -name "*.md" -type f -not -path "*/.git/*" -not -path "*/node_modules/*" | xargs grep -n "^# " 2>/dev/null | head -30

Capture: Doc structure, key files, where to dig deeper.


3. CODEBASE INDEX (How does it work?)

# Entry points
rg "if __name__|^func main" --type py --type go -l 2>/dev/null | head -10

# Classes
rg "^class " --type py -c 2>/dev/null | head -15

# API routes (if applicable)
rg "@(app|router)\.(get|post|put|delete)" --type py -c 2>/dev/null | head -10

Capture: Entry points, core abstractions, API surface.


4. CONFIG SCAN (How is it configured?)

ls pyproject.toml package.json app.yaml cloudbuild.yaml .env* 2>/dev/null
ls .github/workflows/*.y*ml 2>/dev/null | head -5

Capture: Build system, cloud config, CI/CD, environment.


5. GIT STATE (What's in flight?)

git status --short | head -15
git branch -v | grep "^\*"
git log --oneline -5

Capture: Current branch, uncommitted work, recent direction.


6. PROJECT RULES (How should I behave?)

cat CLAUDE.md 2>/dev/null | head -50

Capture: Project-specific instructions, patterns, constraints.


One-Shot Reload (Copy-Paste)

echo "=== SESSIONS ===" && cd ~/.claude/services && python3 conversation-search.py --list-sessions | head -3 && cd - > /dev/null
echo "=== DOCS ===" && find . -name "*.md" -not -path "*/.git/*" | xargs grep -n "^# " 2>/dev/null | head -15
echo "=== CODE ===" && rg "if __name__|^func main" --type py --type go -l 2>/dev/null | head -5
echo "=== CONFIGS ===" && ls pyproject.toml package.json app.yaml .env* 2>/dev/null
echo "=== GIT ===" && git status --short | head -10 && git branch -v | grep "^\*"

After Reload

You now have:

Know Source
What I was working on Conversation history
How the project is organized Doc index
How the code is structured Codebase index
How it's configured Config scan
What's in flight Git status
Project-specific rules CLAUDE.md

Next: Continue previous task OR ask user what's next.


License Corp Context

Repo: lc (licensecorporation/lc)

  • Entity structure: 1-LC, 2-LA, 3-LR, 4-The-Plan, Z-Orchestration
  • Products in x.1, Analytics in x.2
  • FedRAMP High for LR (us-east5 only)

Repo: la (licensecorporation/la)

  • B2B compliance platform
  • Standard GCP regions

Repo: lr (licensecorporation/lr)

  • Government regulatory platform
  • FedRAMP High - us-east5 only

Skill Purpose
repo-map Complete pattern reference
index-docs Doc structure (headings, mermaid, tables)
codebase-index Code architecture
read-configs Config discovery
search-history Conversation search