Session Context Bridge 🔗

Never start from zero. Carry context across sessions with zero dependencies.

The Problem

Every new session starts blind. You lose:

What You Lose	Impact	Recovery Time
Task state & progress	"Where was I?"	5-15 min
Key decisions & reasoning	Re-decide what was already decided	10-30 min
Environment configuration	"Wait, what env vars?"	5-10 min
Blockers & dependencies	Hit same walls again	15-60 min
Project mental model	Re-read codebase	30-120 min

Cost: ~60-240 minutes of re-contextualization per session. At $0.01/token, that's $5-20 wasted per session on re-reading context you already had.

claude-mem (65K★) proved the need. But it requires installation, a worker service, and a database. This skill achieves 80% of the value with 0 dependencies — just markdown files.

The Solution

┌─────────────────────────────────────────────────────────────────┐
│  SESSION END                                                     │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐         │
│  │ current.md   │  │ decisions.md │  │ archive/     │         │
│  │ Active task  │  │ Why we chose │  │ Past sessions│         │
│  │ Next steps   │  │ what we chose│  │ Searchable   │         │
│  │ Blockers     │  │ Alternatives │  │ Reference    │         │
│  └──────────────┘  └──────────────┘  └──────────────┘         │
│       │                   │                  │                  │
│       ▼                   ▼                  ▼                  │
│  ┌─────────────────────────────────────────────────────┐       │
│  │           .context/ directory (git-friendly)         │       │
│  └─────────────────────────────────────────────────────┘       │
│       │                                                         │
│  NEXT SESSION START                                             │
│       ▼                                                         │
│  "Restore my context" → Agent reads .context/ → Picks up      │
│  exactly where you left off                                     │
└─────────────────────────────────────────────────────────────────┘

Quick Start

# Claude Code
cp SKILL.md ~/.claude/skills/session-context-bridge/

# OpenClaw
cp SKILL.md ~/.openclaw/workspace/skills/session-context-bridge/

# Any agent: Copy SKILL.md content into your agent's skill directory

Commands:

• End of session: "save my context"
• Start of session: "restore my context"
• Anytime: "update context" after important decisions

File Structure

.context/
├── current.md        ← What you're working on NOW (auto-updated)
├── decisions.md      ← Architectural decisions with reasoning
└── archive/          ← Previous sessions (searchable)
    ├── 2026-04-24.md
    └── 2026-04-23.md

current.md Template

# Current Context

## Active Task
[What you're working on right now]

## Progress
- [x] Completed step 1
- [x] Completed step 2
- [ ] Next: step 3
- [ ] Then: step 4

## Blockers
- [Any blockers and their status]

## Environment
- Branch: feature/auth
- Last test: PASS (14:30)
- Deploy: staging

decisions.md Template

# Decision Log

## 2026-04-24: Auth Strategy
- **Decision:** Use JWT with refresh tokens
- **Alternatives:** Session cookies, OAuth-only
- **Reason:** Mobile-first, stateless, scalable
- **Trade-off:** More complex token management

Why Zero Dependencies?

Approach	Dependencies	Setup Time	Maintenance
claude-mem	Node.js + Worker + DB	30 min	Ongoing
Vector DB solutions	Embeddings + DB	60 min	Ongoing
This skill	None	30 seconds	None

Markdown files are:

• Git-friendly: Version controlled, diffable, mergeable
• Searchable: grep, ripgrep, any text tool
• Portable: Works with any agent, any OS
• Transparent: You can read and edit them directly
• Zero-risk: No supply chain, no database, no service to maintain

Works With

• OpenClaw
• Claude Code, Cursor, Codex
• Any agent framework

Related Skills

• token-budget-guard — Optimize context window usage
• skill-context-doctor — Diagnose context bloat
• evr-framework — Verify session outputs

License

MIT — Context should persist, not perish.