https://github.com/davdittrich/planning-with-beads
Persistent task & knowledge management for AI agents via Beads system. Based on Manus context engineering principles.
https://github.com/davdittrich/planning-with-beads
agent-skills ai-agents beads claude-extension context-engineering manus persistent-memory planning productivity skill-md task-management
Last synced: 2 days ago
JSON representation
Persistent task & knowledge management for AI agents via Beads system. Based on Manus context engineering principles.
- Host: GitHub
- URL: https://github.com/davdittrich/planning-with-beads
- Owner: davdittrich
- Created: 2026-04-25T15:07:23.000Z (3 months ago)
- Default Branch: master
- Last Pushed: 2026-05-30T22:48:05.000Z (about 2 months ago)
- Last Synced: 2026-05-31T00:19:15.649Z (about 2 months ago)
- Topics: agent-skills, ai-agents, beads, claude-extension, context-engineering, manus, persistent-memory, planning, productivity, skill-md, task-management
- Language: Shell
- Size: 53.7 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Planning with Beads
A persistent task and knowledge management skill for AI agents, based on Manus context engineering principles and powered by the **Beads (`bd`)** system.
## Overview
Planning with Beads replaces volatile file-based planning with a structured, graph-based issue tracker. It ensures that your AI agent maintains a consistent goal, tracks every discovery, and never repeats failed attempts—even across multiple sessions or after context window resets.
## Key Features
- **Hermetic Ticket Architecture**: Enforces execution-ready tickets with explicit scope, input/output schemas, and execution guards.
- **Persistent Memory**: Uses Beads to store tasks, findings, and decisions on disk.
- **Context Engineering**: Implements Manus principles like "Attention Manipulation via Recitation" using `bd prime`.
- **Atomic Workflows**: Enforces one ticket per atomic task and immediate bug tracking.
- **Fail Pattern Prevention**: Includes explicit rules to prevent repetitive mistakes.
- **2-Action Rule**: Ensures every discovery is captured before it's lost from the agent's memory.
- **Automation Scripts**: Helper scripts for session initialization and completion verification.
## Getting Started
1. **Initialize Session**:
The fastest way to start is using the initialization script:
```bash
./scripts/init-session.sh "Your Project Goal"
```
This initializes Beads and scaffolds an Epic with default phases (Discovery, Planning, Implementation, Verification).
2. **Claim and Prime**:
Find an unblocked task and load context:
```bash
bd ready
bd update --claim
bd prime
```
3. **Verify Completion**:
When finished with a session or an Epic:
```bash
./scripts/check-complete.sh
```
## Manus Principles Integration
This skill is designed around the 6 Manus Principles for high-performance AI agents:
1. KV-Cache Optimization
2. Masking vs. Removal
3. Filesystem as External Memory
4. Attention Manipulation via Recitation
5. Persistence of "Wrong Turns"
6. Anti-Few-Shotting Resilience
## Documentation
- [SKILL.md](SKILL.md): Core rules and command reference.
- [reference.md](reference.md): Deep dive into Manus principles with Beads.
- [examples.md](examples.md): Practical workflow examples.
## License
MIT