When to Use This Skill

/auto-summary

Batch-generate one-line summary fields for notes missing this frontmatter property. Uses parallel Haiku agents processing notes in batches of 15-20 for fast, cost-effective summary generation. The summary field enables AI triage, search results ranking, and quick note previews.

When to Use This Skill

• Adding summaries to notes that lack them
• Improving search and discovery across the vault
• Preparing notes for AI-assisted triage and retrieval
• Post-migration cleanup to add summary fields
• Generating previews for index pages or dashboards

Usage

/auto-summary [--scope path/to/folder] [--mode suggest|apply] [--max 100]

Parameters

Instructions

Phase 1: Identify Notes Needing Summaries

1. Scan notes in scope — Find all Markdown files
2. Parse frontmatter — Check for existing summary field
3. Filter: Keep only notes where summary is missing or empty
4. Sort by priority: ADRs and Concepts first (most benefit from summaries), then Projects, then Meetings
5. Divide into batches of 15-20 notes per agent

Phase 2: Batch Summary Generation — Agent Team (Parallel Haiku Agents)

Use the Batch Processing pattern. Launch N parallel agents.

Agent 1-N: Summary Writer (Haiku) Task: Generate one-line summaries for assigned batch

For each note:

1. Read the title, frontmatter, and first 500 words of body content
2. Generate a one-line summary following the Writing Rules and Summary Patterns by Type below
3. Return: List of { filename, generatedSummary } per note

Writing Rules

Every generated summary must follow these constraints:

• Length: 10-25 words, one sentence, no period at the end
• Voice: Start with the subject, not "This note..." or "A document about..."
• Tense: Use present tense for active items, past tense for completed
• Content: Include the most distinctive and searchable terms
• Avoid generic openers: Never start with "Overview of...", "Notes on...", "Summary of...", "Document about...", "Page covering..."
• UK English throughout
• No line breaks — single sentence only

Summary Patterns by Type

Quality Checks

After generating each summary, validate against these rules:

• Too long: Flag if over 25 words — tighten the phrasing
• Too short: Flag if under 10 words — add meaningful context
• Too vague: Flag if the summary could apply to any note of that type (e.g., "Architecture decision about a technology choice")
• Banned phrases: Reject summaries starting with any of these:
  • "This note..."
  • "A document about..."
  • "Summary of..."
  • "Overview of..."
  • "Notes on..."
  • "Page about..."
  • "Document covering..."
• Missing specifics: Flag if the summary lacks any concrete nouns (technologies, processes, outcomes)

If a summary fails quality checks, regenerate it before returning.

Phase 3: Review and Apply

1. Compile summaries from all agents
2. Present for review in a table
3. If --mode apply: Write summaries to frontmatter

Frontmatter format:

summary: "Selected PostgreSQL over MongoDB for transactional consistency in the order domain"

Always quote the summary value since it may contain special YAML characters (colons, brackets).

Output Format

# Auto-Summary Report

**Date:** YYYY-MM-DD | **Notes Processed:** X | **Summaries Generated:** X

## Summaries

| Note                              | Generated Summary                                   | Apply? |
|-----------------------------------|-----------------------------------------------------|--------|
| `ADR - Event Sourcing.md`         | Decision to adopt event sourcing for order audit trail | Yes   |
| `Meeting - 2026-01-15 Sprint.md`  | Sprint planning covering order service backlog priorities | Yes |
| `Concept - CQRS.md`              | Command Query Responsibility Segregation pattern for read/write separation | Yes |

## Statistics

| Note Type  | Processed | With Summary | Without Summary |
|------------|-----------|--------------|-----------------|
| ADR        | X         | X            | X               |
| Meeting    | X         | X            | X               |
| Concept    | X         | X            | X               |
| Project    | X         | X            | X               |
| Other      | X         | X            | X               |

Safety

• Always use --mode suggest first for vault-wide operations
• Never overwrite existing non-null summaries
• Commit to git before running
• If note body is too short (<50 words), skip rather than guess
• Summary generation is additive only — never removes existing summaries

Examples

Example 1: Preview All Missing Summaries

/auto-summary

Example 2: Apply to ADRs

/auto-summary --scope ADRs/ --mode apply

Example 3: Limit to 50 Notes

/auto-summary --max 50 --mode suggest

Invoke with: /auto-summary to batch-generate summaries for notes missing them