cdeust/cortex/cortex-automate

Automate — Triggers, Rules, and Sync

Keywords

trigger, rule, automate, remind me when, auto-remember, sync instructions, push to CLAUDE.md, prospective memory, keyword trigger, file trigger, domain trigger, time trigger, filter rule, boost rule

Overview

Set up proactive automation in Cortex — triggers that fire when conditions are met (like opening a specific file or entering a domain), rules that filter or boost memories during recall, and syncing top insights back into CLAUDE.md for persistent project instructions.

Use this skill when: You want Cortex to proactively surface information based on context, filter out noise during recall, or keep CLAUDE.md updated with memory-derived insights.

Workflow

Triggers — Proactive Memory Recall

Create triggers that fire automatically when conditions match:

Keyword trigger — fires when a query/context contains specific words:

cortex:create_trigger({
  "type": "keyword",
  "pattern": "authentication",
  "memory_id": <id>,
  "message": "Remember: we decided to use JWT with refresh tokens, not sessions"
})

File trigger — fires when a specific file is being worked on:

cortex:create_trigger({
  "type": "file",
  "pattern": "pg_store.py",
  "memory_id": <id>,
  "message": "This file has a known issue with connection pooling under load"
})

Domain trigger — fires when entering a specific project domain:

cortex:create_trigger({
  "type": "domain",
  "pattern": "cortex",
  "memory_id": <id>,
  "message": "Priority: finish the refactoring plan before adding new features"
})

Time trigger — fires after a time condition:

cortex:create_trigger({
  "type": "time",
  "pattern": "7d",
  "memory_id": <id>,
  "message": "It's been a week — run consolidation"
})

Triggers fire up to 5 times by default, then deactivate. They appear in the query_methodology response at session start.

Rules — Filter and Boost During Recall

Add rules that modify recall behavior:

Soft rule (boost/penalize score):

cortex:add_rule({
  "type": "soft",
  "scope": "domain:cortex",
  "condition": "tag:architecture",
  "weight": 1.5,
  "description": "Boost architecture memories in Cortex domain"
})

Hard rule (include/exclude):

cortex:add_rule({
  "type": "hard",
  "scope": "global",
  "condition": "tag:deprecated",
  "action": "exclude",
  "description": "Never surface deprecated memories"
})

Tag rule (auto-tag on store):

cortex:add_rule({
  "type": "tag",
  "scope": "domain:cortex",
  "condition": "content_match:refactor",
  "tag": "refactoring",
  "description": "Auto-tag refactoring memories"
})

List active rules:

cortex:get_rules({
  "scope": "domain:cortex"
})

Sync to CLAUDE.md

Push top memory insights into project CLAUDE.md for persistent context:

cortex:sync_instructions({
  "directory": "<project root>",
  "max_insights": 10
})

This extracts the most important, high-confidence memories and formats them as project instructions in CLAUDE.md. Useful for keeping the project-level instructions file updated with lessons and patterns discovered across sessions.

Tips

• Triggers are personal reminders: Think of them as "note to future self" tied to a context
• Rules shape retrieval: Soft rules are preferred over hard rules — they influence ranking without hiding potentially useful memories
• Sync periodically: Run sync_instructions after significant sessions to keep CLAUDE.md current
• Don't over-trigger: Too many triggers create noise. Use them for genuinely important context-dependent reminders