fireworks-skill-memory

Preserve cross-session lessons and prevent repeated mistakes for coding agents.

93|7|Updated Mar 27, 2026

Search Tags:#claude code #hooks #codex #distillation #skill memory #knowledge files

Ready to run

npx skills add https://github.com/yizhiyanhua-ai/fireworks-skill-memory --skill fireworks-skill-memory

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: fireworks-skill-memory
Source: https://github.com/yizhiyanhua-ai/fireworks-skill-memory/tree/main
Command: npx skills add https://github.com/yizhiyanhua-ai/fireworks-skill-memory --skill fireworks-skill-memory

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

This Skill stops coding agents from starting each session from zero by preserving lessons, mistakes, and workarounds across sessions. It helps prevent the same API bugs, sequencing errors, and configuration pitfalls from repeating endlessly.

Core Features & Use Cases

Persistent per-skill memory keeps lessons scoped to the right skill instead of mixing everything into one global note.
Claude Code uses hook-based context injection and session-end distillation so the agent sees relevant lessons before acting again.
Codex uses explicit inject, checkpoint, and flush commands to load experience, store raw notes, and distill new lessons.
Local knowledge files support cross-session learning for debugging, API usage, workflow consistency, and team-shared skill improvement.

Quick Start

Ask Claude Code to install fireworks-skill-memory from its GitHub repository so the hooks, local memory folders, and Codex command workflow are configured automatically.

Frequently Asked Questions about fireworks-skill-memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I preserve Claude Code lessons across sessions to stop repeated mistakes?▼

Cross-session memory preserves lessons by injecting relevant context via hooks before acting and distilling session-end notes into durable knowledge files. This prevents repeated API bugs and configuration pitfalls by keeping experience scoped per skill.

Can I use skill memory with Codex sessions for context injection and checkpointing?▼

Codex sessions support skill memory through explicit inject, checkpoint, and flush commands. Users load stored experience, capture raw notes during work, and distill new lessons into durable storage to maintain workflow consistency across sessions.

What's the best way to keep per-skill memory scoped instead of mixing everything into one global note?▼

Persistent per-skill memory keeps lessons scoped to the right skill by using local knowledge folders and separate durable files. This ensures debugging notes and API workarounds remain isolated to their specific skill workflows.

Does Claude Code require local hook adapters for session-end distillation?▼

Claude Code requires local hook adapters for context injection and session-end distillation. These hooks automatically configure local memory folders and parse transcripts or summaries to store durable knowledge files for cross-session learning.

How do I install fireworks-skill-memory to configure hooks and local memory folders automatically?▼

Installing fireworks-skill-memory from its GitHub repository automatically configures hooks, local memory folders, and Codex command workflows. This setup enables per-skill memory, context injection, and lesson distillation across coding agent sessions.

Why does my coding agent start each session from zero and repeat the same API bugs?▼

Coding agents start from zero without cross-session memory. Applying per-skill memory with checkpointing and lesson distillation prevents repeated mistakes by ensuring agents see relevant lessons before acting again.