Generate academically bounded fictional personas on demand for multi-agent rehearsal.
Persona Generator stores the grammar for generating personas, not a catalog of predefined personas.
Persona Generator is a local-first, Markdown-first protocol package for creating temporary fictional personas for bounded, low-stakes scenarios. Use this when you want an AI agent to generate a few temporary fictional viewpoints for a low-stakes rehearsal, then stop with an auditable advisory summary.
Start here: Quickstart -> AI discovery guide -> Safety boundaries.
| Persona Generator is | Persona Generator is not |
|---|---|
| A protocol-first Markdown workspace | A hosted service |
| A grammar for on-demand fictional personas | A static persona catalog |
| A bounded construct vocabulary for scenario design | A psychometric instrument |
| A manual-review workflow | A diagnosis tool |
| A dissent-preserving advisory aid | An autonomous decision maker |
| Optional Persona Pack handoff files | Real-person profiling infrastructure |
- Read docs/QUICKSTART.md.
- Ask your AI agent to read AI_DISCOVERY.md for the smallest task-specific path.
- Provide one fictional, low-stakes scenario brief.
- Generate the minimum useful set of ephemeral personas.
- Run the manual review checklist.
- Optionally use the minimal panel adapter only when the task explicitly calls for synthesis.
- Stop with a bounded advisory synthesis.
Copy-paste starter prompt:
Read AI_DISCOVERY.md and docs/QUICKSTART.md. Use Persona Generator only for a fictional, low-stakes scenario. Generate the minimum useful set of ephemeral fictional personas, review them manually, preserve dissent, avoid static persona catalogs, and stop after a bounded advisory synthesis.
Scenario: [describe your fictional low-stakes scenario here]
Decision or rehearsal question: [what should the temporary perspectives examine?]
Why this is fictional and low-stakes: [brief boundary statement]
Desired output: [personas only / personas plus synthesis / optional handoff]
fictional low-stakes scenario
-> academically bounded construct vocabulary
-> LLM-native on-demand ephemeral personas
-> manual review
-> optional minimal panel rehearsal
-> dissent-preserving advisory synthesis
-> optional Persona Pack v0.1 serialization
-> optional handoff
-> stop
Optional panel rehearsal and Persona Pack serialization are task-triggered components. They are not default runtime inventory and they do not create reusable persona presets.
See examples/limited-public-beta for a fictional volunteer knowledge-sharing scenario with multiple ephemeral personas, visible dissent, a preserved minority view, and an advisory synthesis.
- Product decision rehearsal for a fictional beta.
- Social-conflict perspective exploration using fictional roles.
- Creative brainstorming without permanent persona inventory.
- Pre-mortem review for reversible decisions.
- Advisory review for low-stakes plans.
- Optional handoff of one selected fictional persona through Persona Pack v0.1.
All examples are fictional fixtures for format and safety demonstration only. They are not datasets, runtime inventory, reusable presets, or a static persona catalog.
Persona Generator is not for real-person profiling, diagnosis, hiring, surveillance, legal, medical, credit, insurance, educational placement, or other high-stakes decisions.
Read docs/SAFETY_BOUNDARIES.md before adapting the workflow.
Academic constructs are bounded vocabulary for scenario design. They are not psychometric measurements, validation proof, diagnosis, or real-world behavior prediction.
Read docs/ACADEMIC_BOUNDARIES_AND_SOURCE_NOTES.md and docs/EVIDENCE_SUMMARY.md.
- Quickstart
- Architecture
- Academic Boundaries and Source Notes
- Why Persona Generator
- Use Cases
- FAQ
- Safety Boundaries
- AI Discovery Guide
- Experimental llms.txt
Before contributing, read CONTRIBUTING.md, CODE_OF_CONDUCT.md, SECURITY.md, and SUPPORT.md.
Do not submit personal data, sensitive data, covert profiling requests, static persona catalog proposals, psychometric scoring, or high-stakes use cases.
Persona Generator is licensed under Apache-2.0. The license applies to this public repository's docs, protocols, templates, examples, visual assets, and future code unless a file states otherwise.
If this protocol-first approach is useful to you, consider starring the repository so other builders can find it.