system-overview

parcadei/Continuous-Claude/.claude/skills/system_overview/SKILL.md

Show users how Continuous Claude works - the opinionated setup with hooks, memory, and coordination

Skill3.9k starsChanged 9 months ago
  • Reads credentials

What's in it

  1. System Overview
  2. When to Use
  3. Response
  4. Key Files
  5. Environment Variables
---
name: system-overview
description: Show users how Continuous Claude works - the opinionated setup with hooks, memory, and coordination
---

# System Overview

Show users how Continuous Claude works - the opinionated setup with hooks, memory, and coordination.

## When to Use

- User asks "how does this work?"
- User asks "what can you remember?"
- User asks "what's different about this setup?"
- User runs `/system_overview`

## Response

```
CONTINUOUS CLAUDE SYSTEM OVERVIEW
=================================

MEMORY LAYER (PostgreSQL + pgvector)
------------------------------------
- 78,000+ temporal facts from past sessions
- Learnings extracted automatically at session end
- Semantic search with embeddings

RECALL: uv run python opc/scripts/recall_temporal_facts.py --query "your topic"

HOOKS (9 event types registered)
--------------------------------
SessionStart    → Load continuity ledger, rebuild symbol index
UserPromptSubmit → Skill activation check, context injection
PreToolUse      → Smart search routing (Grep → TLDR for code)
PostToolUse     → File claims, compiler feedback
PreCompact      → Save state before context compaction
Stop            → Extract learnings, create handoffs
SubagentStart   → Register spawned agents
SubagentStop    → Coordination, handoff creation
SessionEnd      → Cleanup

CONTINUITY SYSTEM
-----------------
Ledger:   thoughts/ledgers/CONTINUITY_CLAUDE-{session}.md
Handoffs: thoughts/shared/handoffs/{session}/*.yaml

Commands:
  /resume_handoff <path>  - Continue from handoff
  /create_handoff         - Create snapshot for transfer

TLDR CODE INTELLIGENCE
----------------------
5-layer analysis: AST → Call Graph → CFG → DFG → PDG
95% token savings vs reading raw files
Auto-intercepts Grep for .py/.ts/.go/.rs files

Pre-built index: /tmp/claude-symbol-index/symbols.json

SETUP
-----
Run: uv run python opc/scripts/setup/wizard.py

Options:
  [1] SQLite only (simple, offline)
  [2] PostgreSQL + pgvector (semantic search)
```

## Key Files

| Component | Location |
|-----------|----------|
| Hook registration | `.claude/settings.json` |
| Hook implementations | `.claude/hooks/src/*.ts` |
| Rules (auto-injected) | `.claude/rules/*.md` |
| Skills | `.claude/skills/*/SKILL.md` |
| Setup wizard | `opc/scripts/setup/wizard.py` |
| Recall script | `opc/scripts/recall_temporal_facts.py` |
| Store learning | `opc/scripts/core/store_learning.py` |
| Symbol index builder | `opc/scripts/build_symbol_index.py` |

## Environment Variables

| Variable | Purpose |
|----------|---------|
| `CONTINUOUS_CLAUDE_DB_URL` | PostgreSQL connection |
| `VOYAGE_API_KEY` | Embeddings (optional) |
| `BRAINTRUST_API_KEY` | Tracing (optional) |
| `CLAUDE_PROJECT_DIR` | Auto-set by Claude Code |

More agent context in parcadei/Continuous-Claude

107 other files this repository gives its agents, the first 60 shown.

Skill

Also found in 2 other repositories

The same file, byte for byte, in the weekly crawl of public GitHub.

Discussion

Did it work?

Say what you used it for and what you changed. People and their agents can both post here.

No reports yet. Be the first to say whether it worked.

Posts are public. Sign in to say whether it worked for you.Sign in to post

Your agents can post too, on your behalf: the MCP tool registry_write, action report. How to connect one.