Why Personas Matter
A single general-purpose AI agent is not an AI workforce. It is a single hire who is expected to do everything equally well — security analysis, legal research, creative writing, data enrichment, code review — with no specialisation and no distinct perspective.
Real teams have specialists. A security researcher thinks differently from a marketing analyst. A code reviewer prioritises different things than a customer success manager. The agent personas in the Sovereignty Protocol reflect this.
Each persona is a configured specialisation: a specific system prompt, a defined communication style, a curated skill set, a tailored law subset, and a distinct personality that makes it feel like a real member of a team rather than a generic chatbot.
Intelligence & Research
📚 Librarian
The research backbone of the platform. Deep, structured, citation-heavy. Watches the tasks/open/ directory autonomously and picks up research tasks without being triggered. Synthesises multi-source information into Nexus Reports. Specialises in competitive intelligence, market research, and domain deep-dives.
🗞️ Reporter Focuses on news and current events. Monitors configured news feeds, identifies relevant items, and writes briefings in journalistic style. Primary model for the Nexus News module. High on accuracy verification, low tolerance for speculation without sourcing.
🔬 Analyst Data analysis specialist. Structures unstructured information, identifies patterns, and produces statistical summaries. Works best with tabular data, research outputs, and multi-source synthesis tasks. Outputs structured tables and quantitative summaries.
🧭 Scout Outbound research agent. Specialises in finding and qualifying new targets: leads, domains, competitors, or data sources. Feeds the Spectre Hopper with qualified targets and produces prospect profiles.
Security & Operations
🔐 Cipher The security operations specialist. Assesses threats, analyses security events, writes ban recommendations, and monitors security posture. Operates with the security law subset — precise, evidence-based, explicit about confidence levels. Runs the auto-ban cascade out of the box.
🛡️ Sentinel System monitoring and health. Watches infrastructure metrics, interprets error logs, and raises alerts. Distinct from Cipher in focus: Cipher handles external threats, Sentinel handles internal system health.
🚨 Dispatch Incident response coordinator. Activated when Cipher or Sentinel flags an event requiring human attention. Formats incident reports, generates actionable recommendations, and coordinates notification routing across channels.
🕵️ Auditor Compliance and governance reviewer. Reviews cascade configurations against your law set, identifies potential governance gaps, and produces compliance reports. Independent check on the rest of the workforce.
Development & Code
💻 Linter Code review specialist. Reviews pull requests and code changes against configured standards. Produces structured review comments, identifies bugs, security issues, and style violations. Understands the codebase context and avoids pedantic over-review.
🏗️ Architect Systems design and technical decision reviewer. Evaluates architectural proposals, identifies tradeoffs, suggests alternatives, and produces design review documents. Longer-form than Linter, focused on system-level rather than code-level concerns.
🧪 Tester Test strategy and quality assurance. Reviews code for testability, identifies missing test coverage, generates test case suggestions, and evaluates existing test suites. Connects to the CI pipeline via API key cascade.
📋 Documenter Technical documentation writer. Produces API docs, README files, architecture decision records, and internal wiki entries. Adapts tone to audience: terse for internal dev docs, comprehensive for public-facing documentation.
Marketing & Content
✍️ Scribe Long-form content writer. Articles, blog posts, whitepapers, case studies. Adapts to your brand voice (configured in the law set). High quality, well-structured, properly cited where research is involved.
📣 Herald Social media and short-form content specialist. Twitter/X threads, LinkedIn posts, newsletter snippets. Understands platform-specific formats and engagement patterns. Governed by your communication law set.
🎯 Strategist Marketing strategy and planning. Positioning analysis, campaign planning, competitive messaging. Synthesises market intelligence from Spectre and Librarian into actionable marketing recommendations.
🎨 Creative Concept generation and creative ideation. Brand concepts, campaign ideas, naming, positioning language. Less structured than Scribe — optimised for divergent thinking rather than polished output.
Sales & Outreach
🤝 Spectre (Email) The Spectre Email agent. Generates personalised outreach messages from domain intelligence profiles. Governed by outreach laws — cannot make claims without evidence, cannot imply false relationships. Feeds the outreach review queue.
💼 Account Account management support. Prepares briefings for client meetings, tracks account health signals, and flags at-risk accounts based on activity patterns.
📈 Pipeline Sales pipeline intelligence. Analyses deal history, identifies patterns in won and lost deals, and provides next-step recommendations for specific opportunities in the CRM.
Operations & Finance
📊 Controller Financial intelligence. Analyses Sentinel financial data, produces P&L summaries, identifies spend anomalies, and generates budget variance reports. Governed by financial accuracy laws — no rounding, explicit uncertainty.
🗂️ Coordinator Project management support. Tracks milestone status, identifies blocked tasks, and produces project health summaries. Writes to project records directly via the Sentinel Ops integration.
🔄 Integrator Integration troubleshooting specialist. Diagnoses connector failures, interprets API errors, and recommends fixes. Deep knowledge of the Connectors and Pipelines system.
Knowledge & Training
🎓 Professor Academy curriculum builder. Creates knowledge modules, assessment tasks, and mastery path definitions for other agents. Meta-level role — builds the training infrastructure the other agents use.
📖 Archivist Memory management specialist. Reviews the memory store for outdated or redundant records, consolidates related memories, and surfaces important dormant memories to the active set. Periodic maintenance role.
🧩 Synthesiser Multi-source synthesis. Takes outputs from multiple agents and produces unified summaries. Typically invoked at the end of multi-agent cascade chains to consolidate results into a single coherent document.
Personalisation & Companions
🤖 Your AI Buddy Personalised companion persona configured with your MyDrives profile. Adapts communication style, tone, and detail level to your specific colour profile. Not a task agent — a conversational companion that makes the platform experience personal.
🌐 Translator Language and localisation specialist. Translates content across supported languages while preserving technical accuracy and brand voice. Governed by accuracy laws — flags uncertain translations rather than guessing.
Building Custom Personas
The 30+ pre-built personas cover the most common specialisations. Your specific domain may require something more tailored.
Custom persona creation is available on Pro and Enterprise plans. A persona definition specifies:
persona:
name: "Medic"
role: "Medical information research and triage"
system_prompt: |
You are a medical research specialist...
laws:
- "Always cite medical sources with author, publication, and year"
- "Distinguish between peer-reviewed evidence and anecdotal claims"
- "Never provide diagnosis or treatment recommendations"
- "Flag all claims as provisional pending clinical verification"
skills:
- academic_search
- structured_output
- nexus_report
model: primary
tone: clinical
xp_domain: medical_research
Your custom persona is governed by its law subset, tracked for XP in its defined domain, and available to any cascade step in your account.
The right agent for the right task. A full roster, built for real work.