What problem does it solve?
Agents using Friday's memory system often bloat auto-injected system prompts with verbose, redundant entries, lose large analysis outputs when ephemeral artifacts are not referenced, or misuse memory tools across different execution contexts, leading to lost data and inefficient token usage.
Core Features & Use Cases
- Correct Memory Tool Usage: Teaches the proper call patterns for
save_memory_entry, list_memory_entries, and delete_memory_entry across all Friday execution contexts including workspace chat, FSM LLM actions, and Python/TypeScript user agents.
- Terse Entry Formatting: Enforces one-fact-per-entry, under-100-character rules to keep auto-injected memory blocks lean and signal-dense.
- Artifact Reference Pattern: Guides users to store large content (over 500 characters) as durable artifacts with terse memory references, preventing content loss from the 24-hour ephemeral artifact sweeper.
- Use Case: For example, when building a Friday agent that tracks user preferences and stores quarterly sales reports, this skill ensures you write short preference entries to the long-term
preferences store, save full reports as artifacts with memory references, and avoid losing report data after 24 hours.
Quick Start
Use the writing-to-memory skill to save a terse reference to your Q2 sales report artifact in the long-term memory store, delete the outdated "prefer JSON for data exports" entry, and save the new "prefer CSV for data exports" preference entry.
