# Sg Optimize

> Run quality analysis and automated improvement on drafted slide content. Use when you have drafted slides and want to improve quality, optimize readability, or get quality scoring before finalizing.

- Skill: `davistroy/sg-optimize` (Agent Skill)
- Install (CLI): `npx skillmds@latest add davistroy/sg-optimize`
- Raw SKILL.md: https://api.skillmd.com/api/skills/davistroy/sg-optimize/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: davistroy (https://skillmd.com/u/davistroy)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/davistroy/sg-optimize

---


# Slide Generator: Optimize Step

Run quality analysis and automated improvement on drafted slide content. Uses Claude with extended thinking (budget_tokens=4096, temperature=1.0) to score and improve content across 5 dimensions: readability, tone consistency, structure, redundancy, and citation quality.

## Pre-loaded Context

**Working directory:**
!`pwd`

**Available presentation files:**
!`ls -la *.md 2>/dev/null | grep -i pres || echo "No presentation files found"`

## Prerequisites

- **Preflight:** `sg --version` must succeed before proceeding — if `sg` is missing, stop and tell the user this requires the private `davistroy/slide-generator` engine (owner-only; see ADR-0008)
- `slide-generator` package installed
- `ANTHROPIC_API_KEY` set in environment
- A `presentation.md` file from the draft step

## Input Validation

**Required:**
- `<presentation.md>` - Path to presentation markdown file

**Optional:**
- `--output <path>` - Explicit output file path. If omitted, the `sg` engine chooses the output location itself — this is not documented in this repository (the engine lives in the private `davistroy/slide-generator` repo, ADR-0008/D23); run `sg optimize --help` to confirm current behavior before relying on it.

## Instructions

1. **Verify input**: Confirm the presentation markdown exists and contains slide content
2. **Run optimization**:
   ```bash
   sg optimize presentation.md
   ```
3. **Verify output**: Confirm optimized file was created
4. **Report results**: Show quality scores (before/after if available), improvements made

## How It Works

The optimize step uses:
- **Extended thinking**: budget_tokens=4096 for deep analysis (temperature=1.0 required)
- **5-dimension scoring**: Readability, tone, structure, redundancy, citations
- **Representative sampling**: For decks >15 slides, samples representative slides rather than processing all
- **Batched optimization**: Large decks optimized in batches with quality verification
- **Assistant prefill**: Forces structured JSON output for consistent scoring

## Error Handling

| Error | Cause | Fix |
|-------|-------|-----|
| `File not found` | Wrong path | Check file path |
| `No slides detected` | Malformed markdown | Verify presentation.md has proper slide markers |
| `Token limit exceeded` | Very large deck | Set `SG_OPTIMIZE_MAX_TOKENS=32768` or reduce slide count |

## Output

The optimize step produces an improved presentation markdown file — at the `--output` path if given, otherwise wherever the `sg` engine defaults to (unconfirmed here; see `sg optimize --help`) — with:
- Improved readability (shorter sentences, clearer language)
- Consistent tone across all slides
- Reduced redundancy between slides
- Better structural flow and transitions
- Quality score summary

## Next Step

After optimization completes, run graphics validation:
```bash
sg validate-graphics presentation.md
```
Or use `/sg-validate-graphics` to continue the pipeline.