gsd-build

gsd-build / context-packet

Public

File-based context resolution for AI agent DAG workflows. Three primitives, zero dependencies.

18
1
100% credibility
Found Mar 30, 2026 at 21 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

context-packet provides a lightweight, file-based system for defining and executing directed acyclic graph workflows for AI agents, handling context resolution, token budgeting, and state management with simple commands.

How It Works

1
🔍 Find a smart workflow helper

You hear about context-packet, a simple way to chain AI tasks so each step gets just the right info from the ones before, without messy handoffs.

2
📝 Sketch your task chain

You draw a quick map of your project steps, like research leading to outline then draft, noting what each needs from others.

3
📁 Set up your project space

You share your map with the tool, and it makes a cozy folder to keep track of every step's notes and progress.

4
🤖 Team up with your AI friend

You pick your go-to AI helper, like Claude, and connect it so it can jump in and handle each thinking task automatically.

5
▶️ Launch the flow

With one go command, your AI runs through the steps one by one—or in parallel where possible—passing summaries and details smoothly.

6
📊 Watch it unfold

You check the status anytime to see green checks for done steps, knowing exactly what's complete or still brewing.

🎉 Celebrate your output

Your full chain finishes with a neat final result, all context managed perfectly, ready for you to use or share.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 21 to 18 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is context-packet?

Context-packet is a TypeScript library for file-based context resolution in AI agent DAG workflows, using three primitives—init, submit, resolve—with zero dependencies. It solves the wiring nightmare of passing the right upstream context to nodes without manual plumbing: define a graph.json DAG, submit packets when nodes finish, and resolve token-budgeted context with anti-injection wrapping. All state lives in plain .context-packet/ JSON files for easy reset or sharing.

Why is it gaining traction?

Unlike heavy orchestration tools, it ditches databases and servers for dead-simple file storage, letting you run full pipelines via CLI like `context-packet run --agent "claude -p"`. Token budgeting prioritizes summaries and direct deps, idempotency skips unchanged inputs via hashes, and MCP server tools integrate with agents like Claude for stateful sessions. Developers love the low-friction start: drop in graph.json, pipe to any stdin agent, no vendor lock-in.

Who should use this?

AI workflow builders chaining LLMs for tasks like blog generation (research → outline → draft), code review pipelines, or multi-agent systems with parallel nodes. Ideal for TypeScript devs prototyping agent DAGs with Claude, OpenAI, or local models, especially those annoyed by context switching vs packet switching overhead in brittle scripts.

Verdict

Grab it for quick AI DAG prototypes—solid docs and CLI make it dead simple, despite 18 stars and 1.0% credibility signaling early maturity. Test on a toy workflow first; lacks battle-tested scale but shines for solo agent tinkering.

(178 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.