backport-journal

Backport Journal

Import plans from Claude Code's plans directory ($CLAUDE_CONFIG_DIR/plans/, defaulting to ~/.claude/plans/ when the env var is unset) into the Obsidian vault, creating both the vault plan note and a daily journal entry. Plans already imported (tracked via source_slug frontmatter) are skipped. The script resolves the directory itself — no path needs to be passed.

Procedure

1. Discover plans

Run the discovery script from the plugin root directory to list all plans:

bun ${CLAUDE_PLUGIN_ROOT}/hooks/backport-journal.ts --list 2>/dev/null

Parse the JSON output. Each entry has: sourceSlug, title, date, ampmTime, projectLabel, isImported.

Count totals:

• Total plans found in the plans directory
• Already imported to vault
• New (not yet imported)

Tell the user: "Found N plans in the Claude Code plans directory, M already imported to vault, K are new."

If no new plans exist, inform the user and stop.

List the plans still missing, formatted as:

N. [YYYY-MM-DD HH:MM AM] [project/name] Title

2. Ask selection mode

Use AskUserQuestion with these options:

• All new plans — Import all plans not yet in the vault
• Date range — Specify start and end dates to filter
• By project — Filter by project name
• Specific plans — Pick individual plans from the list

3. Gather selection details

If "Date range": Ask the user for start date and end date (YYYY-MM-DD format).

If "By project": Display the unique project labels from the missing plans list. Ask which project(s) to import.

If "Specific plans": Display a numbered list of new plans (format: N. [YYYY-MM-DD HH:MM AM] [project/name] Title). Ask the user which numbers they want (comma-separated). Map their selections back to sourceSlug values.

4. Ask summarization preference

Use AskUserQuestion with these options:

• AI summaries (Claude Haiku) — Richer, more descriptive summaries (uses API calls)
• Fast text extraction — Quick summaries from plan content, no API calls

5. Dry run

Execute the script with --dry-run and the appropriate filters:

# All new plans
bun ${CLAUDE_PLUGIN_ROOT}/hooks/backport-journal.ts --all --dry-run [--skip-summarize] 2>/dev/null

# Date range
bun ${CLAUDE_PLUGIN_ROOT}/hooks/backport-journal.ts --all --from=YYYY-MM-DD --to=YYYY-MM-DD --dry-run [--skip-summarize] 2>/dev/null

# By project
bun ${CLAUDE_PLUGIN_ROOT}/hooks/backport-journal.ts --all --project=project-name --dry-run [--skip-summarize] 2>/dev/null

# Specific plans
bun ${CLAUDE_PLUGIN_ROOT}/hooks/backport-journal.ts --plans=slug1,slug2,slug3 --dry-run [--skip-summarize] 2>/dev/null

Add --skip-summarize if the user chose fast text extraction.

Parse the JSON result and display a summary: "Will import N plans, M will be skipped (already imported), E errors."

If there are errors, show them to the user.

6. Confirm

Use AskUserQuestion to confirm:

• Proceed — Import the plans and create journal entries
• Cancel — Abort without changes

7. Execute

Run the same command without --dry-run:

bun ${CLAUDE_PLUGIN_ROOT}/hooks/backport-journal.ts --all [--skip-summarize] 2>/dev/null

Parse the JSON result and display the final report: "Imported N plans. M skipped. E errors."

If there were errors, list them for the user.