/dex-backlog

name: dex-backlog
description: "Show the AI-ranked backlog of Dex system-improvement ideas. Use when the user says 'show my Dex ideas', 'what's in the backlog', 'what should we build next'. Not for workshopping one idea into a plan; use `dex-improve`. Not for discovering existing features; use `dex-level-up`."

What This Command Does

**In plain English:** AI-powered ranking of your Dex system improvement backlog based on current system state. Shows you what to build next.

**When to use it:**

• Weekly check-in on system improvements
• After capturing several new ideas
• When deciding what to work on next
• During quarterly planning for system improvements

**How to run it:**

/dex-backlog              # Full review with re-ranking

---

Process Overview

1. **Load context** - Read system state, usage patterns, learnings 2. **Score ideas** - Calculate 5-dimension scores for each idea 3. **Re-rank backlog** - Sort by weighted total score 4. **Update file** - Write new rankings to `System/Dex_Backlog.md` 5. **Present top ideas** - Show top 5 with "Why now?" justification 6. **Offer next steps** - Workshop, implement, or defer

---

Step 1: Load System Context

Read these files to understand current system state:

Required Files

System/Dex_Backlog.md              # All ideas to score
System/usage_log.md                # Feature adoption patterns
System/user-profile.yaml           # Role, preferences
CLAUDE.md                          # Current capabilities

Optional Files (if they exist)

System/Session_Learnings/           # Recent pain points (last 30 days)
.claude/commands/                  # Available commands
core/mcp/                          # MCP integrations
06-Resources/Learnings/               # Captured patterns

Extract Context

Build a context dictionary with:

• **Usage patterns**: Which features are used vs. unused
• **Role profile**: PM, Sales, Leadership, Engineer, etc.
• **Pain points**: Recent friction from session learnings
• **System capabilities**: What's currently available
• **Backlog state**: All ideas and their current scores

---

Step 2: Score Each Idea

For every active idea in the backlog, calculate 5 dimension scores.

⚠️ CURSOR FEASIBILITY CHECK (Do This First!)

Before scoring ANY idea, validate it's actually implementable in Cursor:

**What Cursor/Terminal CAN do:**

• ✅ Read and write files
• ✅ Execute shell commands
• ✅ Build MCP tools for structured operations
• ✅ Parse and transform file contents
• ✅ Create caches and indexes (file-based)
• ✅ Run commands on schedules or triggers

**What Cursor/Terminal CANNOT do:**

• ❌ Track user edits in real-time
• ❌ Hook into Cursor internals
• ❌ Monitor user actions passively
• ❌ Access edit history without explicit file reads
• ❌ Real-time background processes watching for changes

**If idea requires something from the CANNOT list → Set all scores to 0 and flag as "Not feasible in Cursor"**

---

After feasibility check passes, score on 5 dimensions:

Dimension 1: Impact (35% weight)

**Question:** How much would this improve daily workflow?

**Scoring logic:**

Base score: 50

+20 if matches_recent_pain_points():
  - Search System/Session_Learnings/ for mentions of this issue
  - Keywords from idea title/description appear in learnings
  - Problem stated explicitly in recent notes

+15 if affects_daily_workflow():
  - Touches commands used >3x per week (from usage_log)
  - Modifies core files (03-Tasks/Tasks.md, daily plans, person pages)
  - Impacts repetitive actions

+15 if has_compound_value():
  - Enables other ideas in backlog
  - Reduces technical debt
  - Creates reusable patterns
  - Unblocks multiple workflows

Max: 100

**Examples:**

• "Auto-suggest person pages": 95 (daily workflow + enables relationship tracking)
• "Export to blog": 40 (nice-to-have, doesn't affect core workflow)

---

Dimension 2: Alignment (20% weight)

**Question:** Does this fit actual usage patterns?

**Scoring logic:**

Base score: 50

+30 based on usage_overlap():
  - Extract features idea depends on
  - Check if those features are used (usage_log)
  - Calculate overlap: (used_features / total_features) * 30

+20 if fits_role_profile():
  - PM roles: prioritize project/product features
  - Sales roles: prioritize relationship/account features
  - Leadership: prioritize synthesis/review features
  - Match category to role focus areas

Max: 100

**Examples:**

• Idea needs "person pages" → Check if user has created person pages
• If usage_log shows person pages = used → Higher alignment
• If role = PM and idea = product features → +20 role fit

---

Dimension 3: Token Efficiency (20% weight)

**Question:** Does this reduce context/token usage?

**CRITICAL - Cursor Feasibility Check:** Before scoring, verify the idea is implementable in Cursor/Terminal:

• ✅ Can use: File read/write, MCP tools, command execution, file-based caching
• ❌ Cannot use: Real-time edit tracking, Cursor internal hooks, monitoring user actions
• If not feasible in Cursor → Score = 0 on all dimensions

**Scoring logic:**

Base score: 50

+25 if reduces_token_usage():
  - Caches/stores frequently accessed data (in files/MCP)
  - Compresses or summarizes verbose content (file-based)
  - Eliminates redundant reads
  - Enables more efficient retrieval patterns

+15 if improves_context_efficiency():
  - Reduces number of files that need reading
  - Creates structured summaries (YAML/JSON files)
  - Better indexing/search to avoid broad scans
  - Moves data from markdown to structured format

+10 if enables_incremental_updates():
  - Supports partial updates instead of full rewrites
  - Tracks changes in separate files
  - Lazy loading or on-demand computation

Max: 100

**Examples:**

• "Cache meeting summaries in YAML": 90 (file-based, avoids re-reading)
• "Track user edits for learning": 0 (NOT FEASIBLE - can't track edits)
• "Add new field to template": 50 (neutral token i