Putting Your Memory to Work: Context-Aware Recipes
Beyond Simple Macros
Traditional automation tools ask you to define every step, and they break the moment the work changes shape. The interesting version is different: the AI agent you already use — Claude, Cursor, Codex — reads your memory and works from what actually happened.
Context, not schedules
The useful prompts are not "run this at 9am every day." They are questions that only make sense against a record of your day:
- After every standup — "recap what we agreed and who owns what"
- While reviewing a PR — "what did we decide about this module last month, and where"
- Before a client call — "what has changed on this account since we last spoke"
Trying your first one
Getting started is simple:
- Open the agent you already work in
- Ask it a question about your own week
- It reads your LUCI memory for the parts it needs
- You get an answer with the moment it came from
Multi-step questions
The good ones chain. "Find the thread where the scope changed, then draft the note to the client" needs both the record and the reasoning, and your agent has both once it can read your memory — no state machine to maintain, no steps to wire together.
Recipes worth stealing
A few that come up constantly:
- Meeting prep — the agenda, the last conversation, and the open threads
- Code review — the decision behind the code, not just the diff
- Weekly report — what you actually worked on, across every app
Coming Soon
We are collecting the prompts people find themselves reusing, so good recipes spread instead of being reinvented.
