attache-analyst

name: attache-analyst description: | This skill should be used when analyzing OmniFocus system structure, discovering workflow patterns, running AI-enhanced GTD coaching, or learning the user's tool stack. Triggers when user asks "analyze my OmniFocus", "discover my system", "what's my structure", "AI coaching", "project health analysis", "system insights", "learn my workflow", "understand my patterns", "suggest improvements", "what tools do I use", or "update my tool stack".

Do NOT trigger for: "show tasks" or "check inbox" (use omnifocus-core), "create OmniFocus plugin" (use omnifocus-generator), "what makes a good next action" (use gtd-coach for pure methodology).

This skill is read-only against OmniFocus. It discovers patterns, infers structure, and persists learned insights via the Attache preferenceManager. It can read and update vault tool notes (status, usage patterns) via the archivist agent. license: MIT metadata: version: 1.0.0 author: totally-tools compatibility: platforms: [macos] requires: - OmniFocus 4.8+ (for Foundation Models actions) - macOS 26+ with Apple Silicon (for on-device AI)

Attache Analyst

System learning, pattern detection, and AI-enhanced coaching for the Attache agent.

Capabilities

System Discovery

Discover the user's OmniFocus structure using hybrid rule-based and Foundation Models inference.

# Install/update the Attache plugin
open assets/Attache.omnifocusjs

# Trigger system discovery (writes System Map to task note)
SCRIPT='var p=PlugIn.find("com.totallytools.omnifocus.attache");var lib=p.library("systemDiscovery");var map=lib.discoverSystem({depth:"full"});var t=flattenedTasks.filter(function(t){return t.name==="Attache System Map"});if(t.length>0){t[0].note=JSON.stringify(map);Pasteboard.general.string="System Map updated v"+map.attacheVersion}else{Pasteboard.general.string="Task not found"}'
open "omnifocus://localhost/omnijs-run?script=$(python3 -c "import urllib.parse,sys;print(urllib.parse.quote(sys.stdin.read().strip()))" <<< "$SCRIPT")"

System Map fields:

• tags.categories.contexts[] — GTD context tags
• tags.categories.people[] — waiting/delegation tags
• tags.categories.status[] — someday/on-hold tags
• tags.categories.duration[] — duration/effort tags
• tags.categories.schedulingContext[] — time-of-day/week tags
• structure.topLevelFolders[] — folder names and inferred types
• tasks.dataQuality.percentWithDuration — derive durationModel

Pattern Inference

The TypeScript libraries in scripts/src/attache/ provide:

Tool Stack Awareness

Read the vault's Tools.base to discover the user's full tool landscape:

• Parse tool categories, statuses (Use/Trial/Assess/Hold), and host relationships
• Build a tool registry mapping tools to agents
• Update tool note usage patterns when observed

Access via archivist agent — do not read vault files directly.

Attache Plugin (On-Device AI)

The Attache.omnifocusjs plugin provides 9 on-device AI actions:

• Daily and weekly GTD reviews
• Project analysis and health scoring
• System discovery with persistent memory
• Completed task summaries

Build: bash scripts/build-attache.sh

Boundaries

• Read-only against OmniFocus data — never create, update, or delete tasks
• Persist learned patterns only via preferenceManager (non-invasive)
• Defer task mutations to omnifocus-core skill
• Defer plugin creation to omnifocus-generator skill

Reference Documentation

• references/foundation_models_integration.md — Apple Intelligence API details
• references/library_ecosystem.md — Attache library architecture and overlap guide
• references/insight_patterns.md — GTD health scoring patterns