---

name: automations-memory description: "Use and troubleshoot Khoj scheduled automations, cron normalization, scheduler leadership, and user memories." disable-model-invocation: true metadata: disco-role: operating license: AGPL 3.0

Khoj Automations and Memory

Use this sub-skill when working on Khoj scheduled automations, /api/automation, /api/memories, cron normalization, background scheduler leadership, user memory settings, or memory scoping by user and agent.

Route Map

• Use automations-api.md for /api/automation CRUD, manual trigger behavior, required parameters, cron normalization, response metadata, and /automated_task query prefixing.
• Use memory-api.md for /api/memories behavior, update-as-delete-and-recreate semantics, memory settings, and user/agent scoping rules.
• Use scheduler-runtime.md for APScheduler, DjangoJobStore, process-lock scheduler leadership, timezone handling, and asynchronous trigger execution.
• Use troubleshooting.md for invalid cron strings, minute-level recurrence rejection, duplicate automations, missing conversations, scheduler leader conflicts, missing memories, and disabled or mis-scoped memory.
• Use validate_cron.py to safely validate and normalize Khoj-style cron strings without starting the server or touching a database.

Boundaries

• This sub-skill owns scheduled automation APIs, cron normalization, automation runtime scheduling, scheduler leader behavior, memory APIs, memory settings, and memory user/agent scoping.
• Route ordinary chat payload construction, slash-command response generation, model selection, and chat agent behavior to chat-agents.
• Route server startup, database migrations, deployment, authentication setup, Docker, and general REST route orientation to deployment-api.
• Route search, content retrieval, indexing, file filters, embeddings, and reranking to search-retrieval or content-indexing.

Evidence Basis

This guidance is distilled from Khoj source, docs, and tests for automation routing, memory routing, scheduler setup, process locks, memory settings, and scheduled-chat actor behavior. It is self-contained for future coding agents and does not require reading original repository files at runtime.