Myriad

Generates structured prompts by composing from a library of 35+ pre-tested rules and 150+ instructions organized by content type (articles, ads, email, scripts). Users select applicable rules (e.g., 'click-worthy titles', 'power words', 'target audience specification') and the system assembles them into a cohesive prompt instruction set. Rules are tested specifically against ChatGPT's behavior but claimed compatible with Copilot, Gemini, Claude, and Llama. The system detects rule conflicts and allows priority marking with '!' to enforce precedence when contradictions arise.

Takes existing content (article, ad, email, etc.) and rewrites it according to selected rules from the library. The system applies transformations to enforce style, tone, keyword integration, call-to-action directives, and audience targeting without requiring manual prompt construction. Users specify which rules to apply and the tool generates a prompt that instructs the backend LLM to rewrite while adhering to those constraints. Output is generated via copy-paste workflow to external LLM services.

Applies audience-targeting rules that enforce content generation for specific demographic, psychographic, and behavioral audience segments. Rules guide the backend LLM to use language, examples, and references appropriate for the target audience (e.g., 'Gen Z', 'B2B executives', 'small business owners'). The system generates prompts that specify audience characteristics and tested against ChatGPT's ability to tailor content appropriately. Rules include audience persona definitions, language preferences, and cultural references.

Allows users to define custom rules beyond the predefined library of 35+ rules and add them to their personal rule library for reuse. Custom rules are stored and can be applied to future prompts alongside predefined rules. The system supports custom rule composition, naming, and description. Custom rules are not shared across users and are not validated against predefined rules for conflicts. Custom rules are treated identically to predefined rules in prompt generation and conflict detection.

Exports generated prompts in formats suitable for sharing, copying, and reusing across team members and external LLM services. Prompts are exported as plain text formatted for copy-paste into ChatGPT, Copilot, Claude, Gemini, and Llama interfaces. The system supports exporting individual prompts or collections of prompts for a content type. Exported prompts include all selected rules, instructions, and metadata. No programmatic API export or structured format (JSON, YAML) is documented.

Analyzes existing competitor or reference content to extract underlying patterns, rules, and structural elements that make it effective. Users input competitor content and the system generates a prompt that instructs an LLM to decompose the content and identify the rules, tone, structure, and techniques used. Results are returned as a structured analysis that can inform new prompt creation. This enables reverse-engineering of successful content patterns without manual analysis.

Identifies contradictions when multiple rules are selected simultaneously (e.g., 'formal tone' vs 'casual tone', 'long-form' vs 'concise'). The system flags conflicting rules and allows users to mark priority rules with '!' to enforce precedence when contradictions arise. This prevents generating prompts that contain mutually exclusive instructions that would confuse backend LLMs. The conflict detection is rule-aware and based on the predefined rule library's known incompatibilities.

Generates prompts optimized for multiple backend LLM services (ChatGPT, Microsoft Copilot, Google Gemini, Claude, Llama) from a single rule set. The system claims to adapt the same rules across different model APIs, though documentation indicates primary optimization for ChatGPT with compatibility claims for others. Users select their target LLM and the system generates a prompt formatted for that service's API or interface. No direct API integration is provided — prompts are generated for manual copy-paste into each service.

Generates a secondary prompt that instructs the backend LLM to verify whether its previous output actually followed the specified rules. The system creates a verification prompt that asks the LLM to review its own output against each rule and provide evidence (examples) that it complied. This enables quality assurance without manual review, though verification accuracy depends on the LLM's self-assessment capability. The verification prompt is generated separately and requires manual execution against the LLM's previous output.

Provides curated rule sets optimized for specific content types: long-form articles, listicles, ads, email, scripts, webpages, and social media. Each content type has pre-selected rules and instructions tailored to that format's conventions and best practices. Users select a content type and the system presents only relevant rules, reducing decision complexity and improving output quality. Rules are tested specifically for each content type's requirements (e.g., email rules focus on subject lines and CTAs, article rules focus on structure and SEO).

Applies tone and style-of-voice rules from the library to ensure consistent brand voice across generated content. Rules include options like 'formal', 'casual', 'professional', 'conversational', 'authoritative', etc. The system generates prompts that enforce specific tone constraints and can detect conflicts when incompatible tones are selected (e.g., 'formal' + 'casual'). Tone rules are integrated with other content rules and tested against ChatGPT's instruction-following capability for tone consistency.

Applies SEO-focused rules that enforce keyword integration, keyword density, semantic keyword inclusion, and search-intent alignment in generated content. Rules guide the backend LLM to naturally incorporate target keywords without keyword stuffing, maintain appropriate keyword density, and include related semantic keywords. The system generates prompts that specify keyword targets and integration constraints, tested against ChatGPT's ability to balance SEO requirements with readability.

Applies CTA-focused rules that enforce specific call-to-action directives, CTA placement, CTA clarity, and conversion-focused language in generated content. Rules guide the backend LLM to include explicit CTAs, use action verbs, create urgency, and align CTAs with content goals (e.g., 'sign up', 'download', 'buy now'). The system generates prompts that specify CTA requirements and tested against ChatGPT's ability to write persuasive, conversion-focused CTAs.