estafeta lets coding agents carry one project across many chats, agents and machines: each new session resumes from the project's own files, not from chat history.

In practice:

• You open a chat with "Use session-start". The agent finds where work stopped, checks what changed since, and continues.
• Work runs in phases with acceptance criteria. A phase is DONE only after a check has actually passed.
• "Use session-end" saves the exact next step for whoever continues: another chat, another agent or another machine.

You need an agent with terminal access (Claude Code, Codex, Gemini CLI and others) and Python 3.9+.

Install estafeta from https://github.com/estafeta-dev/plugin
for this agent: follow "Agent installation" in docs/INSTALL.md.
I authorize its installer to install and replace the estafeta skills
in this agent's skills directory, with backups.

The last sentence is your consent for the agent to change its own skills. Without it, a guarded agent such as Claude Code in auto mode stops before install.

git clone https://github.com/estafeta-dev/plugin.git estafeta
cd estafeta
python3 scripts/workflow.py install --all-detected

Then ask any agent: "Use session-guide. Explain how this works."

Say	When
Use session-start	Opening a chat
Use plan-and-orchestrate	Work longer than one step
Use agent-relay	Ask, review with or delegate to another AI provider
Use session-handover	At a milestone
Use session-end	When you stop

Installation · Usage · Benchmarks · Validation

---

This project is licensed under the Apache License 2.0.