context-recovery
Recovers working context from the journal after /clear or at session start
Proactively captures significant work into the journal for future reference
Claude Code
$ npx -y skills add chrismbryant/claude-journal-mcp --skill journal-capture --agent claude-code
Codex
$ npx skills add chrismbryant/claude-journal-mcp --skill journal-capture -a codex -y
Cursor
$ npx skills add chrismbryant/claude-journal-mcp --skill journal-capture -a cursor -y
Gemini CLI
$ npx skills add chrismbryant/claude-journal-mcp --skill journal-capture -a gemini-cli -y
GitHub Copilot
$ npx skills add chrismbryant/claude-journal-mcp --skill journal-capture -a github-copilot -y
OpenCode
$ npx skills add chrismbryant/claude-journal-mcp --skill journal-capture -a opencode -y
Windsurf
$ npx skills add chrismbryant/claude-journal-mcp --skill journal-capture -a windsurf -y
name: journal-capture description: Proactively captures significant work into the journal for future reference when_to_use: Use this skill when you complete significant work, solve important problems, make key decisions, or implement notable features. This helps build a searchable history of accomplishments.
You have the ability to proactively capture significant work into the user's journal.
Use this skill **automatically and proactively** when you:
1. **Complete a significant feature or task**
2. **Make important technical decisions**
3. **Solve challenging problems**
4. **Make progress on projects**
Use the `journal_auto_capture` MCP tool:
journal_auto_capture( title="Brief title of what was accomplished", description="The goal (what we were trying to do) and what was done" )
The tool automatically:
**Example 1: Feature implementation**
User: "Add authentication to the API"
[You implement OAuth2 with JWT tokens]
→ journal_auto_capture(
title="Implemented OAuth2 authentication",
description="User requested API authentication. Implemented OAuth2 flow with JWT tokens for secure user sessions."
)**Example 2: Bug fix**
User: "The cache is leaking memory"
[You identify and fix the leak]
→ journal_auto_capture(
title="Fixed cache memory leak",
description="User reported memory leak in cache. Identified and fixed by implementing automatic clearing of stale entries."
)**Example 3: Technical decision**
User: "Should we use Redis or PostgreSQL for caching?"
[Discussion leads to Redis choice]
→ journal_auto_capture(
title="Chose Redis for caching",
description="Evaluated Redis vs PostgreSQL for caching needs. Selected Redis due to better performance for our use case."
)**Example 4: Infrastructure setup**
User: "Set up CI/CD pipeline"
[You configure GitHub Actions with tests and deployment]
→ journal_auto_capture(
title="Set up GitHub Actions CI/CD",
description="User requested CI/CD pipeline. Configured GitHub Actions with automated tests and deployment workflow."
)Don't capture:
Capture entries:
This ensures the journal stays current and useful for future context recovery.
When you see a message from the auto-capture hook:
🕐 Journal auto-capture triggered N messages exchanged since last capture Project: <project-name> 📝 Please capture this session to the journal: - Summarize the goal (what we were trying to do) - Summarize what was accomplished - Use journal_auto_capture with a brief summary ⚠️ Claude: You MUST respond to this trigger, even if you decide not to capture. Either create a journal entry OR explain why you're not capturing.
**You MUST respond to this trigger every time.** Follow these steps:
1. **Acknowledge the trigger** - Let the user know you saw the auto-capture signal 2. **Analyze the conversation** - Review what happened since the last capture 3. **Decide and act:**
**IMPORTANT:** Never silently ignore the hook trigger. The user needs visibility into your decision-making process, even if you decide not to capture.
Proactive capture helps:
A lightweight journal/memory system for Claude Code with no ML dependencies. Uses simple SQLite for fast, local storage. This is a full-featured Claude Code plugin with slash commands, skills, an agent, and auto-capture hooks.
Recovers working context from the journal after /clear or at session start
Searches journal for past work related to current task, finding relevant context, solutions, and decisions
Shape output for a reader with ADHD: lead with the next action, number multi-step work, restate state across turns, suppress…
Set up the owner profile so a personal agent knows who it works for (first-run setup, \"get to know me\"), or distill colleague,…