
checkpoint-session
by rollercoaster-dev
Monorepo for Rollercoaster.dev OpenBadges ecosystem - shared packages, applications, and documentation
SKILL.md
name: checkpoint-session description: Manage session lifecycle, learning extraction, and context metrics. Use when starting/ending sessions, analyzing workflows for learnings, querying knowledge graph, or checking session metrics. allowed-tools: Bash, Read
Checkpoint Session Skill
Manage Claude session lifecycle, learning extraction, and context metrics.
When to Use
- Starting a new Claude session (auto-run via SessionStart hook)
- Ending a session and extracting learnings
- Analyzing completed workflows for patterns and mistakes
- Querying the knowledge graph for relevant learnings
- Checking session metrics and usage patterns
CLI Reference
All commands use the checkpoint CLI:
bun run checkpoint <command> [args...]
Session Lifecycle Commands
Session Start
Initialize session context and check for resumable workflows:
bun run checkpoint session-start [--branch <name>] [--issue <number>]
What it does:
- Detects current git branch and modified files
- Cleans up stale workflows (>24 hours)
- Prompts for workflow resume if active workflows found
- Outputs session context summary
Session End
Record session metrics and extract learnings:
bun run checkpoint session-end [options]
Options:
--workflow-id <id>- Associate with specific workflow--session-id <id>- Session identifier (auto-detected from session-start)--learnings-injected <n>- Count of learnings used in session--start-time <iso>- Session start timestamp--compacted- Flag if context was compacted--interrupted- Flag if session was interrupted--review-findings <n>- Count of review findings--files-read <n>- Count of files read
Learning Commands
Analyze Workflow
Extract learnings from a completed workflow:
bun run checkpoint learning analyze <workflow-id> <dev-plan-path>
Analyzes:
- Planned vs actual commits
- Deviations from plan
- Review findings and fixes
- Patterns to repeat
- Mistakes to avoid
Query Learnings
Find relevant learnings from the knowledge graph:
bun run checkpoint learning query [options]
At least one filter required:
--code-area <area>- Filter by code area (e.g., "authentication", "database")--file <path>- Filter by file path--issue <number>- Filter by source issue number
Returns:
- Learning content
- Confidence scores
- Related patterns
- Related mistakes
Metrics Commands
List Session Metrics
View recorded session metrics:
bun run checkpoint metrics list [issue-number]
Shows per session:
- Files read
- Compaction status
- Duration
- Review findings
- Learnings injected/captured
Metrics Summary
View aggregate metrics:
bun run checkpoint metrics summary
Shows:
- Total sessions
- Compaction rate
- Average files read
- Average learnings injected/captured
- Total review findings
Example: Full Session Flow
# 1. Session starts (auto-run by hook)
bun run checkpoint session-start --branch feat/issue-394
# 2. During session: use /work-on-issue or /auto-issue
# These use the workflow checkpoint commands
# 3. After workflow complete: analyze for learnings
bun run checkpoint learning analyze "workflow-394-xxx" ".claude/dev-plans/issue-394.md"
# 4. Session ends (capture metrics)
bun run checkpoint session-end --files-read 42 --review-findings 3
Example: Query Knowledge Before Starting Work
# Find learnings about similar code areas
bun run checkpoint learning query --code-area "badge-verification"
# Find learnings from a related issue
bun run checkpoint learning query --issue 380
# Find learnings about a specific file
bun run checkpoint learning query --file "packages/openbadges-types/src/ob3.ts"
Integration with Hooks
The session commands integrate with Claude Code hooks:
| Hook | Command | Purpose |
|---|---|---|
| SessionStart | session-start | Inject learnings, check workflows |
| SessionEnd | session-end | Extract learnings, record metrics |
Session metadata is automatically passed between hooks via temp files.
スコア
総合スコア
リポジトリの品質指標に基づく評価
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
3ヶ月以内に更新がある
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
レビュー
レビュー機能は近日公開予定です