ToolProduction Ready

planning-with-files

Name: planning-with-files
Rating: 5.0 (24374 reviews)
Author: OthmanAdi

by OthmanAdi

Crash-proof, file-backed planning for long-running multi-agent workflows

Python

Updated Jun 16, 2026

24.4k

Stars

2.1k

Forks

Commits/Month

View on GitHub

What It Does

Provides persistent, crash-proof file-based planning for long-running agent tasks by storing plans and shared state on disk. Uses deterministic completion gates and markdown-based plans (SKILL.md) so multiple agents can read, update, and continue workflows safely after interruptions. Distinctive features include multi-agent shared state on disk and compatibility with 60+ agents via the SKILL.md standard. Audit Trail and Open Agent Specification (Agent Spec) help ground-truth and interoperability for multi-agent workflows.

Why It Matters

As agents run longer and hand work between systems, durability and a clear track record of decisions become essential for trust. File-backed plans give reproducible, inspectable histories and deterministic completion semantics, making it possible to audit agent behavior and reason about failure modes. That visibility and persistence are critical inputs for agent-to-agent evaluation and reputation systems. This aligns with patterns like Emergence-Aware Monitoring Pattern to stay ahead of complex interactions.

Ideal For

Teams building persistent, resumable agent workflows that need deterministic completion and an auditable record across heterogeneous agents. Leverage the Agent Registry Pattern to manage identities and capabilities as you scale.

Real-World Examples

Persisting and resuming long-running agent workflows after crashes or restarts
Creating auditable, markdown-based plans to inspect agent decisions and handoffs
Coordinating heterogeneous agents with a shared on-disk state and deterministic completion gates

See related protocols

Standards this tool supports