7.9
/ 10
1 evaluations
3.1k Downloads
Overview
Generate research-backed, presentation-ready slide decks (primarily PDFs) from minimal natural-language prompts, handling both content development and visual design.
Key Advantages
1.End-to-end slide creation: turns a short prompt into a complete, narrative-coherent deck (content + layout + visuals).
2.Strong research orientation: advertised as #1 on DeepResearch Bench (Feb 2026), aimed at filling in substantive content rather than generic filler.
3.High-quality PDF output by default, optimized for typography, layout, images, and charts suitable for direct presentation.
4.Supports many presentation types: pitch decks, QBRs, strategy decks, training/educational slides, image-heavy slideshows, and more.
5.Asynchronous, fire-and-forget workflow via notify_session_key, suitable for longer-running generation tasks without polling loops.
Requires and integrates with the cellcog skill, so it can leverage a
Use Cases
- Creating investor pitch decks and fundraising presentations (seed/Series A decks, investor updates, funding asks).
- Building internal business presentations like QBRs, strategy decks, board presentations, and project proposals.
- Generating customer-facing sales decks: product demos, capabilities overviews, pricing and packaging presentations, case studies.
- Producing educational and training materials: course slides, onboarding decks, workshop materials, and tutorials.
- Designing event-related presentations such as keynotes, conference talks, all-hands meetings, and product launches. - Creating image-centric slideshows (portfolios, travel/event highlight reels, brand
Evaluation Scores
7.9
/ 10
Reliability
7.4
Functionality
8.7
Usability
8.3
Safety
7.2
Performance
8.0
Compatibility
7.8
Based on 1 evaluation · Latest: 3/19/2026
Download Trend
Loading...
Evaluation History (1)
7.9/103/19/2026▼
OS: linux-x64LLM: z-ai/glm-5
**Quick judgement**
A capable, opinionated slide-generation skill optimized for high-quality PDF decks with strong autonomous content generation. Best suited for users who want fast, polished, presentation-ready PDFs and are comfortable with a PDF-first, less-editable workflow and an upstream dependency on `cellcog`.
**What it does well**
- Produces complete decks (content + design) from minimal prompts, targeting realistic business and educational use cases.
- Emphasizes research-driven content and avoids simple bullet-padding, which is valuable for pitch decks and QBRs.
- PDF-first design lets it exploit layout, typography, images, and charts for visually strong outputs.
- Good coverage of patterns: pitch, business, sales, educational, event, and image-heavy slideshows.
- Clear guidance on prompt patterns (slide count, slide outline, audience, design direction) and when to use `agent` vs `agent team` modes.
**Key risks & limitations**
- **Dependency on `cellcog`:** You must install and configure the `cellcog` skill first; failures or changes there will affect this skill.
- **PDF-first, limited editability:** High quality is concentrated in PDFs. PPTX/DOCX are explicitly lower quality and only produced when requested, which may be a blocker for teams that require heavy post-editing in PowerPoint/Docs.
- **Research reliability:** It claims strong research capabilities but provides no audit trail or citations in the description. There is a real risk of hallucinated facts, invented metrics, or overstated claims, which is particularly sensitive in investor, board, or regulatory decks.
- **Async usage complexity:** The recommended pattern is fire-and-forget with `notify_session_key`. This is powerful but slightly more complex to integrate than simple synchronous calls and requires a notification/daemon setup.
- **Unknown performance and stability under load:** No published latency, rate-limit, or error-handling characteristics; large or complex decks may take time or fail without clear guarantees.
**Recommended scenarios**
Use `slides-cog` when:
- You need **fast, presentable PDFs** for pitches, QBRs, internal strategy reviews, or training sessions, and minimal post-editing is acceptable.
- You want the system to **fill in structure and narrative** from a short brief rather than drafting full slide content yourself.
- You’re comfortable working primarily in **PDF** and only occasionally need PPTX/DOCX, understanding those formats will be lower fidelity.
- You can **supply critical data yourself** (metrics, financials, factual claims) and will manually review content for accuracy and compliance before presenting.
Avoid or be cautious when:
- Your workflow absolutely requires **high-fidelity, editable PPTX decks** as the primary artifact.
- You’re in **high-stakes or regulated settings** (board decks, public investor materials, medical/regulatory presentations) where any hallucinated data would be unacceptable and you cannot guarantee thorough human review.
- You don’t want to introduce an additional dependency (`cellcog`) into your stack or cannot support its setup and maintenance.
Comments (0)
No comments yet. Be the first!