# Memory Proactive

> Proactive layered recall and generic domain-aware routing.

- Skill: `mage0535/memory-proactive` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add mage0535/memory-proactive`
- Raw SKILL.md: https://api.skillmd.com/api/skills/mage0535/memory-proactive/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: mage0535 (https://skillmd.com/u/mage0535)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/mage0535/memory-proactive

---


# Memory Proactive

This skill describes the current proactive recall surfaces that remain useful in
the local repository.

## Layered Context Injection

Primary engine: `tiered_context_injector.py`

It combines multiple recall paths before the next interaction:

- recent session recall from shared SQLite state
- governance-backed object and hub recall
- knowledge note recall
- optional live Hindsight recall when the query profile needs it

Example:

```bash
python3 scripts/tiered_context_injector.py --query "user preferences"
```

Useful knobs:

- `--min-score 0.3`
- `--max-results 10`
- `--recall "project status" "system config"`
- `--domains project,stock`

## Generic Domain Routing

Helper: `domain_memory.py`

This is a lightweight legacy helper for splitting local memory into generic
domains so one topic does not crowd out everything else.

Example domains:

- `project`
- `stock`
- `system`
- `marketing`
- `relationship`
- `general`

Example:

```bash
python3 scripts/domain_memory.py check project "Milestone planning note"
python3 scripts/domain_memory.py status
```

## Guidance

- use generic domains, not user- or project-specific hardcoded labels
- keep recall queries portable across agent runtimes
- treat this skill as an optional helper layer, not a required production
  contract