suggest

<arc_log_context>
**Use Read tool:** `.arc/log.md` (first 50 lines)

Check what was recently worked on to avoid re-suggesting completed work.
</arc_log_context>

# Suggest Workflow

Analyze tasks, codebase, and vision to give opinionated recommendations for what to work on next.

## Priority Cascade

1. **Existing tasks** (highest priority) — Already noted, most immediate
2. **Codebase issues** — Technical debt, gaps, patterns
3. **Vision gaps** — Goals not yet implemented
4. **Discovery** (lowest priority, opt-in) — New feature ideas from external research

## Process

### Step 1: Check Tasks

**Use TaskList tool** to check for existing tasks.

If tasks exist with status `pending`:
→ Recommend those first with brief rationale

### Step 2: Analyze Codebase

**Use Task tool to spawn exploration agent:**
```
Task Explore model: haiku: "Analyze this codebase for:
- Incomplete features (TODOs, FIXMEs)
- Technical debt (outdated patterns, missing tests)
- Quality issues (type escapes, inconsistencies)
- Missing documentation
- Performance concerns

Prioritize by impact."
```

### Step 3: Read Vision (if needed)

Only if no tasks exist AND codebase analysis found nothing urgent:

**Use Read tool:** `docs/vision.md`

Compare vision goals to current state. Identify gaps.

### Step 4: Synthesize Recommendations

Present top 3-5 suggestions:

```markdown
## Suggestions

### 1. [Top recommendation]
**Why:** [Brief rationale]
**Command:** /arc:ideate [topic]

### 2. [Second recommendation]
**Why:** [Brief rationale]
**Command:** [relevant command]

### 3. [Third recommendation]
**Why:** [Brief rationale]
**Command:** [relevant command]
```

### Step 5: Offer Discovery

**After presenting normal suggestions**, if fewer than 3 strong recommendations emerged from Steps 1-3, or if the suggestions are mostly maintenance (tech debt, TODOs), offer discovery:

"Those are the immediate priorities. Want me to look beyond the codebase — research what's trending in your space and propose n