Memory & Agents Prompts

Deciding what to remember, what to forget, and how agents hand work to each other. 16 prompts in this category.

• Memory Write Policy DesignUse once, early. The output becomes the write policy you encode in the agent.
• Conflicting Fact ResolutionFor the specific conflicts your store surfaces, before you hard-code a global rule.
• Agent Tool-Use PlanningGood for designing the loop before you build it, and for debugging why an agent picks the wrong tool.
• Memory Retrieval Query RewriterWhen recall is poor because users phrase questions differently than they phrased the original fact.
• Context Budget AllocatorUse when you are hitting context limits and need a truncation order that is not just 'drop the oldest'.
• Agent Failure Post-MortemThe first thing to run on any agent failure, since the fix differs entirely by failure class.
• Memory Summarization Without LossFor compacting old sessions without destroying the details a later query will need.
• Multi-Agent Handoff BriefUse in pipelines where the second agent otherwise re-derives work the first already did.
• Personalization Signal ExtractionRun periodically over history to seed memory without storing every turn.
• Memory Staleness DetectorFor auditing an aging memory store before its stale entries start producing wrong answers.
• Agent Loop Termination ConditionsBefore an agent runs unattended, to prevent infinite loops and runaway cost.
• Tool Description OptimizerWhen the agent is calling the wrong tool or passing bad arguments.
• Memory Conflict Merge RulesFor building a memory store that does not silently overwrite important history.
• Agent Self-Correction TriggerFor agents that need to recover from bad tool outputs without human intervention.
• Episodic vs Semantic Memory SplitterWhen your store distinguishes memory types and you need clean initial data.
• Agent Goal DecomposerWhen a single agent is doing too much and you want to split responsibilities cleanly.

Browse every category →