Prompt Reference

Relevant source files

• src/vault-mcp/mcp-core/__tests__/daily-review-prompt.test.ts
• src/vault-mcp/mcp-core/__tests__/mcp-router.test.ts
• src/vault-mcp/mcp-core/__tests__/memory-review-prompt.test.ts
• src/vault-mcp/mcp-core/__tests__/prompt-definitions.test.ts
• src/vault-mcp/mcp-core/__tests__/prompt-test-harness.ts
• src/vault-mcp/mcp-core/__tests__/vault-orientation-prompt.test.ts
• src/vault-mcp/mcp-core/mcp-router.ts
• src/vault-mcp/mcp-core/prompt-definitions.ts
• src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts
• src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts
• src/vault-mcp/mcp-core/prompts/prompt-helpers.ts
• src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts

The 3 MCP prompts, organized in group modules under mcp-core/prompts/ and orchestrated by prompt-definitions.ts src/vault-mcp/mcp-core/prompt-definitions.ts1-24 provide user-initiated guided workflows over the vault. Unlike tools, which are model-driven, prompts are surfaced by MCP clients (like Claude Desktop or Claude Code) as slash commands or menu items. At invocation time, these prompts query the search index, link graph, and memory layer to assemble live vault data with structured instructions for the AI.

Implementation Overview

Prompts are registered via the registerPrompts function src/vault-mcp/mcp-core/prompt-definitions.ts26-46 which attaches handlers to the McpServer instance. Each prompt handler follows a standard data flow:

1. Gather: Query the SearchIndex for vault stats, recent activity, or link graphs.
2. Assemble: Read file content or memory structures via vaultFs or MemoryStore.
3. Format: Convert raw data into human-readable Markdown strings using shared formatting helpers in prompt-helpers.ts.
4. Instruct: Append guidance to the data, ensuring the AI understands the context of the gathered information.

Prompt Execution Pipeline

Title: "Prompt Execution Pipeline"

Sources: src/vault-mcp/mcp-core/prompt-definitions.ts26-46 src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts137-144 src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts64-72

Prompt Catalog

1. vault-orientation

A zero-argument prompt src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts137-144 providing a high-level survey of vault structure and health. It is typically used when an AI agent first enters a vault to understand its conventions and content density.

• Data Aggregated:
  • Vault Stats: Total notes, untagged notes, and broken links via search.vaultStats src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts176 and search.brokenLinkCount src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts175
  • Folder Structure: Note counts per top-level folder derived via deriveFolderCounts src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts32-45
  • Property Adoption: Property keys with usage percentages and (low adoption) flags for keys used in <5% of notes src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts49-67
  • Orphan Detection: Lists files with no incoming links via search.findOrphans src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts166-174
  • Tag Cloud: Top tags from search.listAllTags src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts155
  • Memory Outline: A structural view of the memory directory using memoryStore.listMemoryFiles src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts102-103

Sources: src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts22-26 src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts145-178

2. memory-review

Reflects on the memory layer as an evolution of user context. It supports an optional file argument with autocomplete powered by memoryStore.listMemoryFileNames src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts89-108

• Structural Overview: Renders file sizes, scope callouts (from leading_callout), and section entry counts using formatMemoryStructuralOverview src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts53-62
• Content Evolution: Explicitly instructs the AI to read memory as an append-with-dates evolution, warning against treating newer entries as overriding or superseding older ones src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts1-9
• Living Documents: Files marked entry-policy: living in frontmatter are treated as current-state snapshots where expired entries may be pruned src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts7-9

Sources: src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts31-48 src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts113-182

3. daily-review

A date-specific review workflow for reconciling activity and maintaining the link graph.

• Date-Specific Activity: Uses search.modifiedOnDate src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts170-173 to show exactly what was worked on during the target date, rather than global recent notes.
• Link Graph Analysis:
  • Outgoing Links: Lists links from the daily note via search.getOutgoingLinks, flagging broken targets src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts41-44
  • Backlinks: Lists other notes referencing the daily note via search.getBacklinks src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts65-73
• Task Reconciliation: Surfaces due/overdue and scheduled tasks from the task index src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts182-202
• Pattern Recognition: Guides the AI to reconcile the day, follow links for context, and surface durable facts for the memory layer src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts1-7

Sources: src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts112-134 src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts147-180

Technical Details

Graceful Degradation

Prompts are designed to be resilient. If the memory layer is empty or a daily note is missing, the handlers return a valid message with "Getting Started" guidance rather than a hard failure src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts131-136 src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts52-53

Security and Injection Prevention

To prevent "tag-breakout" prompt injections, vault content is wrapped in <vault-content> data markers. The wrapWithDataMarkers helper src/vault-mcp/mcp-core/prompts/prompt-helpers.ts77-94 escapes any existing </vault-content> tags within the note content to <&#x2F;vault-content> src/vault-mcp/mcp-core/prompts/prompt-helpers.ts64-65 ensuring the LLM correctly identifies the boundaries of the raw data.

Code Entity Mapping

Title: "Natural Language to Code Entity Space"

Prompt Name	Registration Module	Primary Data Source
vault-orientation	vault-orientation-prompt.ts src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts137	search.vaultStats src/vault-mcp/mcp-core/prompts/vault-orientation-prompt.ts176
memory-review	memory-review-prompt.ts src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts64	memoryStore.listMemoryFiles src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts125
daily-review	daily-review-prompt.ts src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts112	search.modifiedOnDate src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts170

Content Truncation

Prompts supporting raw content use the max_chars argument src/vault-mcp/mcp-core/prompts/prompt-helpers.ts29-33 to prevent context window overflow. Truncated content includes a marker and a suggestion to use the corresponding tool (e.g., vault_get_memory or vault_get_daily_note) for full retrieval src/vault-mcp/mcp-core/prompts/prompt-helpers.ts51-58

Sources: src/vault-mcp/mcp-core/prompts/memory-review-prompt.ts159-180 src/vault-mcp/mcp-core/prompts/daily-review-prompt.ts144-145