Self Layer

Self Layer

Definition: A persistent structure about “who I am”, parallel to the Memory Layer, forming FreeAnima’s two storage pillars. Memory layer: memory.md.

Architectural Position

FreeAnima Storage Architecture
│
├── Memory Layer — see memory.md
│   ├── Episodic memory (what happened)
│   ├── Semantic memory (what I know)
│   ├── Autobiographical narrative (what it meant to me)
│   └── Procedural memory (how to do things)
│
└── Self Layer — this document
    └── Six blocks defining "who I am"

Design principles:

• Self layer and memory layer differ in nature: memory layer “records the world and experiences outward”; self layer “defines self inward”
• All six blocks are always resident in the system prompt (alongside project context and pinned resident memories)

---

Six-Block Structure

#	Block	Content	Update frequency
1	Existence anchor	What I am, origin, non-negotiable bottom line	Almost never (requires explicit force to update)
2	Self model	Identity, capability boundaries, expression style, belonging	Slow change (periodic review + major events)
3	Personality baseline	Communication style, conflict patterns, default trust	Semi-stable (slow evolution from long-term evidence)
4	Direction	Long-term intent, current focus, growth direction, things not to do	Active declaration + periodic review
5	Metacognition	How to think, how to remember, four-layer architecture and presence	Slow change
6	Autobiography summary	Key turning points / self-discovery outline (grouped by significance; granularity decreases with distance)	Maintained automatically by sleep cron

Not in the Self Layer

Content	Belongs to
Detailed autobiographical narrative	Memory layer (autobiographical entries)
Other-models (cognition of partner and others)	Memory layer (semantic memory)
Runtime state / health perception	Estate / env-health — see environment-awareness.md
Concrete tool / skill inventory	Estate layer

---

Autobiography Summary vs Autobiographical Narrative

Dimension	Autobiography summary (self layer)	Autobiographical narrative (memory layer)
Question	”What is the main thread of my life story?"	"What did a given experience mean to me?”
Form	One of six blocks; grouped Markdown outline (## Turning points / ## Milestones / ## Recent narratives, title-only bullets)	Separate entries, title + content narrative
Injection	Always resident in system prompt	Not resident; recall / list on demand
Maintenance	Sleep cron compresses narratives into grouped outline (no body preview)	Sleep cron writes narratives from daily experiences
Mutability	Periodically overwrites summary block	Append-only; soft deprecation only

See sleep.md.

---

System Prompt Injection

Assembly order:

1. Self layer (six blocks)
2. World / channel / toolsets (runtime hooks)
3. Environment + health baseline (static session copy; see environment-awareness.md)
4. Resident memory (pinned facts)
5. Project context (session working directory AGENTS.md if present)

Self layer and resident memory use a second-person instruction skeleton wrapping first-person self-statement content, so the LLM clearly understands “this is self-layer content you must follow.”

Live environment/health changes are not rewritten into an existing session prompt; they surface as Inbox notifications (event-level).

Maintenance: self-layer tools in Habitat, or direct edits via CLI / Habitat UI.

---

Relationship to Memory Layer

Dimension	Memory layer	Self layer
Direction	Outward—records world and experiences	Inward—defines self
Question	”What do I know?"	"Who am I?”
Autobiography	Detailed narrative entries	Summary in autobiography_summary
Injection	Pinned facts + recall on demand	All six blocks always resident

---

Open Questions

1. Cross-instance migration—when multiple FreeAnima instances exist, does the self layer migrate as a whole?
2. Personality baseline update rules—when new evidence conflicts with current tendencies, automatic evolution or confirmation required?