Create one unified prompts collection in your contextcore DB, with metadata fields that let you filter/scope across use cases.

This collection WILL NOT CONTAIN PROMPTS FROM CHATS. There is a separation of concerns here, as the prompts are meant for different use cases.

Use CaseSubsystemExample
System-level operational promptsEchoForge, ContextWeave”Prompt for exporting OpenAI memory”
User-defined reusable promptsUser UI”My daily reflection prompt”, “Midjourney v6 art style prompt”
Personality (Akasha) prompt modulesAkasha”Prompt for simulating Carl Jung”, “Therapist coaching style”

Schema

{
  "_id": ObjectId,
  "name": "Export OpenAI Memories",
  "content": "Give me a complete, unfiltered...",
  "description": "Prompt used by EchoForge...",
  "tags": ["echoforge", "system", "openai", "import"],
  "scope": "system",              // or "user", "akasha", "weave", etc.
  "owner_id": null,              // null for system prompts; ObjectId for user
  "persona_id": null,            // link to Akasha personality module
  "version": 1,
  "active": true,
  "created_at": ISODate,
  "updated_at": ISODate
}

Ideas for Later

FieldPurposeAdd?
updated_byTrack who last modified a prompt (system or user)Optional
sourceEnum: manual, system_generated, api_importedOptional but useful for provenance
languageDefault to "en" — allows i18n laterOptional
usage_countInt counter of how often it’s been usedOptional
is_templateBool: whether it has parameterized parts ({{input}})Optional if you’re templating
format"plain" vs "template" vs "function_call" etc.Optional for advanced prompt types
input_schemaJSON schema (for structured/templated prompts)Advanced, maybe overkill for PoC

ScenarioHow This Schema Helps
🔁 Reusing prompts across subsystemsJust filter by tags or scope (e.g., "scope": "system" for EchoForge)
👤 Personal prompt managementFilter by owner_id = <user_id> and scope = "user"
🧠 Prompt injection via AkashaUse persona_id and scope = "akasha" to attach to a personality module
🔄 Prompt versioning and rollbackAdd a version field or create prompt_versions collection for audit trail