fireworks-tech-graph
Proof: Technical diagrams as durable product and distribution infrastructure.
The Lab / working proofs
Repositories are answers under test.
Every repository answers a question that is still open. Read the questions first, then the proofs.
- Can per-skill memory stop coding agents from repeating the same mistakes?
- Does session persistence make long-running agent work resumable instead of fragile?
- Can architecture diagrams become a durable publishing artifact, not decoration?
fireworks-skill-memory
Proof: Per-skill memory that turns past work into reusable operating context.
fireworks-sessions-saver
Proof: Checkpointing for coding sessions that will inevitably be interrupted.
paper-pal
Proof: Research reading as an active workflow of return, notes, and synthesis.
media-downloader
Proof: A utility layer for turning media gathering into a repeatable workflow.
fireworks-open-eli5
Proof: Evidence-aware interactive visual explainers for Codex and Claude Code.
skills-updater
Proof: Maintenance infrastructure for living skill ecosystems.
fireworks-design
Proof: 🎆 A Claude Code workflow that replicates ClaudeDesign — parallel design exploration, panel judging, synthesis, and adversarial refinement to a single world-class frontend page. 双语 README。
chuinb-skill
Proof: A narrow skill as evidence of taste in workflow packaging.
youtube-ai-digest
Proof: A workflow for turning noisy AI video streams into digestible signals.
zimage-skill
Proof: Image generation as a reusable workflow component, not a one-off trick.
agent-skills
Proof: A skill library for turning repeated prompts into reusable operating units.
fireworks-macapp-creator
Proof: A bridge from agent workflows into packaged desktop software.
fireworks-ai-tools-memory
Proof: Memory for tool-using AI systems that need continuity across work.
skill-knowledge-extractor
Proof: A tool for turning tacit workflow know-how into extractable assets.
fireworks-radio
Proof: Ambient audio as an AI-native working surface.
skill-adoption-planner
Proof: A planning layer for whether a skill should actually be adopted.
rhythm-master
Proof: A playful interface that still proves interaction design range.
agent-harness-engineering-zh
Proof: 三份独立的 Agent Harness 中文知识文档、PDF、阅读站与人工确认周更系统
fireworks-half-vibe
Proof: AI writes the scaffolding; you write the soul. A Claude Code / Codex skill that keeps your real programming skill alive during vibe coding — five composable modes, zero homework vibes. AI 写脚手架，灵魂你来写。
llm-inference-engineering-zh
Proof: 19 篇 LLM 推理工程资料的中文主题化知识重构、阅读网站与人工确认周更系统
fireworks
Proof: Official website for 一支烟花 AI 社区
agent-architect-lab
Proof: A public lab for agent-system structure, not just agent demos.
kv-agent-lab
Proof: A lab for key-value state and durable memory patterns in agent systems.
agent-shared-memory-demo
Proof: A concrete window into shared memory for multi-agent work.
skill-roi-calculator
Proof: An ROI lens for deciding whether an AI skill is worth operationalizing.
.github
Proof: The operating surface for the whole organization.