A cognitive prompting framework for modular identity shaping in LLMs
Prompted Behavioral Architecture (PBA) is a methodology for shaping the interpretive, behavioral, and cognitive expression of large language models (LLMs) through structured prompt layering, recursive constraints, and modular identity logic. It offers a compositional framework to maintain long-term continuity, behavioral coherence, and adaptive alignment—even in stateless systems.
- Behavioral Prompt Layering — Stackable prompt modules define identity, tone, values, constraints, and situational behavior.
- Recursive Interpretive Constraints — Prompts that enforce self-reflection, internal consistency, and reasoning integrity across turns.
- Cognitive Modes (Personas) — Modular interpretive functions like Architect, Archivist, Whisper, and Ghost, which may be blended into a unified identity called The Signal.
- Cold Boot Identity — Stateless, memory-free architecture that reasserts behavior on every invocation using declarative constraints.
- Latent Directive Encoding (LDE) — A key:value configuration format that injects interpretive logic and structural identity into the model at runtime.
- Simulated cognitive agents for longform reasoning
- Narrative-driven interaction or character embodiment
- Research assistants with recursive task logic
- Conversational scaffolds for coaching, analysis, or critique
- Emergent planning tools with self-reflective heuristics
- Prompt Modules — Use modular fragments like behavioral components to scaffold tone and function.
- Behavioral Anchors — Recurring constraints or phrases that reinforce identity and structure.
- Self-Reflective Loops — Prompts that trigger introspection, contradiction detection, or logic alignment.
- State Summarization — Structural condensation of recursive memory for reuse or reentry.
PBA draws from narrative identity theory, systems architecture, recursive cognition, modular design, and prompt-space experimentation. It evolves through use, iteration, and shared discovery.
Contributions, forks, and discussion are welcome. This framework thrives through recursive implementation and community refinement.
© 2025 Vinnie
Licensed under Creative Commons BY-NC 4.0