Skip to content

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 candidate to active to stable
  • 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

  1. Review new candidates in Memory Inbox and approve high-value items.
  2. Edit confidence or content where needed.
  3. Create one or more context modules (for example: API work, debugging, release checks).
  4. Preview generated packs and adjust selectors.
  5. Use generated packs in agent workflows.

Guides

API Surfaces

  • GET/POST/PUT/DELETE /api/v1/memory-items
  • GET /api/v1/memory-items/search
  • GET /api/v1/memory-items/global
  • POST /api/v1/memory-items/{id}/promote
  • POST /api/v1/memory-items/{id}/deprecate
  • POST /api/v1/memory-items/merge
  • POST /api/v1/memory-items/extract/preview
  • POST /api/v1/memory-items/extract/apply
  • GET/POST/PUT/DELETE /api/v1/context-modules
  • POST /api/v1/context-packs/preview
  • POST /api/v1/context-packs/generate

Use the full API schema in skillmeat/api/openapi.json for request and response details.