Memory & Context Intelligence Guide¶
SkillMeat's Memory & Context Intelligence System helps you capture project knowledge, review it quickly, and build reusable context packs for future agent runs.
What You Can Do¶
- Triage candidate memories in a keyboard-first inbox (
/projects/[id]/memory) - Promote important memories from
candidatetoactivetostable - Deprecate stale or incorrect memories with reason tracking
- Build reusable context modules with selector rules
- Preview and generate token-budgeted context packs
- Run memory workflows from CLI (
skillmeat memory ...) - Extract candidate memories from run logs (
extract preview/apply)
Core Concepts¶
- Memory item: A single piece of project knowledge (
decision,constraint,gotcha,style_rule,learning) - Lifecycle status:
candidate,active,stable,deprecated - Context module: A named set of selector rules for building context
- Context pack: The generated output that combines selected memories within a token budget
Typical Workflow¶
- Review new candidates in Memory Inbox and approve high-value items.
- Edit confidence or content where needed.
- Create one or more context modules (for example: API work, debugging, release checks).
- Preview generated packs and adjust selectors.
- Use generated packs in agent workflows.
Guides¶
API Surfaces¶
GET/POST/PUT/DELETE /api/v1/memory-itemsGET /api/v1/memory-items/searchGET /api/v1/memory-items/globalPOST /api/v1/memory-items/{id}/promotePOST /api/v1/memory-items/{id}/deprecatePOST /api/v1/memory-items/mergePOST /api/v1/memory-items/extract/previewPOST /api/v1/memory-items/extract/applyGET/POST/PUT/DELETE /api/v1/context-modulesPOST /api/v1/context-packs/previewPOST /api/v1/context-packs/generate
Use the full API schema in skillmeat/api/openapi.json for request and response details.