Intelligent Slide Designer

The Problem It Solves

Crafting professional presentations extends far beyond filling text into templates. It requires balancing three core challenges: precisely conveying the user's intent, persuading a specific audience visually, and translating the design into a high-quality PPT file through efficient, reliable engineering. Traditional tools or generic approaches often falter in one area—for instance, overlooking audience analysis, relying on generic templates that yield bland content, or sacrificing quality due to an inefficient design-build-check cycle.

How It Works

slide-maker is engineered to operate like an experienced senior designer. Its core methodology centers on treating user requirements as the absolute source of truth and enforcing a rigorous, phased process to ensure design quality and execution efficiency.

Core Principles & Key Process

1. Demand Exploration & Strict Fidelity: The skill begins with an "interview," not action. It proactively asks questions to understand the user, audience, and purpose. All provided materials (old PPTs, templates, documents) are treated as raw material to be processed. It adheres strictly to the source content, prohibiting any fabrication or embellishment, ensuring every data point and claim is traceable.

2. The "Taste Protocol": This is the core design philosophy. It mandates that the agent first judge like a person ("What would the sharpest art director in this room do here, and why?") at every decision point before validating with machine rules (checklists). The goal is to produce a deck with "signature moves" and discernment, not just a "correct but dead" template-driven product that passes compliance checks.

3. Efficient Engineering: The skill emphasizes matching effort to stakes. It leverages toolchains like python3 scripts/dispatch_brief.py and deck_cycle.py, uses the --example flag for rapid component reference, and parallelizes independent tasks. Key habits include: writing a deck brief once and referencing it everywhere; using Edit instead of rewriting entire build scripts for fixes; and iterating efficiently via deck_cycle.py to avoid redundant tool calls and context re-sending overhead.

4. Built-in Critic Loop: Any final output undergoes quality assurance. The skill features built-in critic loops of varying intensity: from a default quick scan post-build, to two focused "lens" critics (content & design), to a multi-critic and arbiter panel for high-stakes projects. This loop can never be skipped by the agent on its own authority; its termination is solely decided by the user during final review.

Operating Modes & Quality Control

The skill offers two modes:
* Standard Mode: The agent autonomously handles the interview, sets checkpoints (🔴 CHECKPOINT), builds, and runs the critic loop. Critical stages pause for mandatory user confirmation.
* Collaborative Mode: Engages the user for approval at key decision points (e.g., style direction, outline), suitable for brand-defining projects.

All design decisions and checkpoint states are recorded (e.g., in .deck-gates.json), ensuring transparency and traceability. The built-in LOOP BREAKER mechanism prevents meaningless repair loops, enforcing a fundamental redesign based on measurement when the same fault persists.

Scope & Caveats

• User Final Authority: The critic loop is mandatory, but its termination depends on the user's decision after reviewing the final product. The agent cannot shut down this flow itself.
• Mode Auto-Waiver: The "decide everything" directive only enables an auto-waiver for the current deck. Starting a new project defaults back to the standard checkpoint flow, preventing accidental loss of control.
• Generation, Not Pure Creation: The skill focuses on optimizing design and engineering implementation based on existing user content and requirements, not on creating from pure concept. Forward-looking content (e.g., a "future work" slide) is only permitted when explicitly needed and can be extrapolated correctly.
• Dependence on Toolchain: The skill's efficient building and iteration are deeply reliant on its preset scripts (e.g., python3 scripts) and workflow. It is designed as a professional tool, not a simple instruction follower.