Agent skill
journal-manager
Creates and manages Obsidian work journals. Automatically writes journals after experiments, analyses, and key decisions. Periodically reviews, consolidates, archives, and promotes journal entries to permanent documentation.
Install this agent skill to your Project
npx add-skill https://github.com/fubira/claude-code-settings/tree/main/skills/journal-manager
SKILL.md
Journal Manager Skill
Create and manage Obsidian work journals. Record "what I thought at this point" as thinking logs.
Activation Triggers
Create (proactive): After experiments/analyses, decisions, incident responses, comparisons, or user request
Organize: /journal-review, /journal-cleanup, auto-suggest when active files > 20
Journal Location
- Path:
/mnt/c/Users/matsushita/obsidian/notes/WORK/{ORG}_{PROJECT}/journal/YYYY-MM-DD_HHmm_topic.md - Date in JST. Topic in hyphenated English or Japanese (keep short)
auto memory vs Journal
- auto memory: Technical facts and patterns (long-lived) →
.claude/projects/*/memory/ - Journal: Chronological thinking process (short-to-mid lived) → Obsidian
journal/
Part 1: Creating Journals
Templates
Experiment/Analysis: Background → Conditions/Setup → Results (numeric tables) → Findings → Conclusion/Next actions Decision: Context → Options → Judgment and reasoning → Trade-offs → Next actions Incident Response: Situation → Root cause → Actions taken → Prevention measures → Lessons learned Work Log: Tasks done → Design decisions (if any) → TODOs (remaining)
Principles
- One topic per file (multiple per day OK). Use tables for numeric data
- Record "what I thought at this point" — valuable even if conclusions change later
- Place persistent data in the parent directory of
journal/ - Do NOT record: code diffs (Git handles that), trivial task logs, technical facts suited for auto memory
Part 2: Review and Organization
Phase 1: Review (analysis only, no changes)
- Read all files with Explore agent, classify by theme
- Determine status for each file:
- Archive: Conclusion already reflected elsewhere / superseded by later work / completed TODOs / approach replaced
- Consolidate: 3+ consecutive files on same theme / problem-identification + solution pairs
- Defer: Explicitly marked "future work" research topics (with start conditions)
- Keep: Sole source of information / latest analysis on active theme / incident response records
- Present organization proposal as a table, wait for user approval
Phase 2: Organize (after approval)
- Archive: Move original file to
archives/as-is (no content changes) - Consolidate: Read source files → create merged file (oldest date + theme name, list sources at top, preserve numeric tables faithfully) → move originals to
archives/ - Defer: Merge related files into one in
deferred/(state resume conditions at top) → move originals toarchives/
Phase 3: Promotion Check
| Detected Pattern | Promote To |
|---|---|
| Confirmed technical facts/patterns | MEMORY.md |
| Project convention changes | CLAUDE.md |
| Model performance records | profile.md |
| General-purpose solutions | knowledge/ |
After promotion, add "→ reflected in X" marker to the journal entry.
Directory Structure
{project}/journal/
├── *.md # Active
├── archives/ # Consolidated/obsolete originals
└── deferred/ # Future research topics
Important Notes
- Archive means move, not delete (originals kept in archives/)
- Never drop data during consolidation (faithfully copy numeric tables)
- When in doubt, keep. Create aggressively, organize cautiously
- Deferred entries must state concrete resume conditions (not just "someday")
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
refactoring-assistant
Assists with code refactoring by detecting code smells, suggesting improvements, and providing refactoring patterns. Activates when writing/editing code, explicitly requested refactoring, or when code quality issues are detected. Maintains awareness of core principles while providing detailed patterns and examples.
code-reviewer
Assists with code review by analyzing code changes for quality, best practices, security, and potential issues. Activates after implementing code features, bug fixes, or refactorings. Provides structured feedback with critical issues, suggestions, and positive highlights.
context-compactor
Analyzes and compacts context-affecting documents (project memory, CLAUDE.md, skill files) to reduce token usage and compaction frequency. Manual trigger only. Always requires user approval before making changes.
test-executor
Executes tests, analyzes test results, checks test coverage, and provides comprehensive testing status overview. Primarily for Go projects. Activates after implementing/modifying code to verify correctness, or when explicitly requested to assess test suite health.
knowledge-manager
Manages a structured knowledge base of patterns, troubleshooting guides, best practices, and workflows. Activates when discovering reusable insights, solving technical problems, or establishing new standards. Records knowledge in categorized files for future reference without bloating global CLAUDE.md.
release-assistant
Automates and ensures reliable release workflows with automatic version bump based on commit history, mandatory lint/build/test execution before release, and safe tag creation and push.
Didn't find tool you were looking for?