- Home
- Skills
- Workflow Automation
- Multi-Agent Ledger
Multi-Agent Ledger
Stop two AI agents overwriting each other: claims, a task log and handoffs in one LEDGER.md.
Free
Works with the AI Agents you already use
Multi-Agent Ledger
Example session with this skill installed
Two agents keep clobbering each other in this repo. Set up a ledger.
- Read your context and instructions
- Compiled the multi-agent ledger
Created LEDGER.md at the repo root from the template, filled Current state from the README and git log, and added "Before any work, read LEDGER.md and follow its protocol. Update it before you stop." to the top of CLAUDE.md (after asking). Open it any time: Claims shows who is working on what; lines marked needs: human are questions for you. It's an agreement between agents, not a lock. To stop using it, delete LEDGER.md and that one line.
Connects securely to your tools. The creator never sees your data.
About this skill
One shared file every agent reads first and updates last. Run parallel subagents, two Claude Code windows, or Claude plus Codex or Cursor in one repo without them editing the same files.
What you get
- Advisory claims: one writer per set of files, with a written rule for who wins an overlap (earlier UTC start time; ties go to the row higher in the table).
- A task log with commit hashes and how each change was verified, so the next agent knows what happened.
- A backlog, locked decisions, and open issues tagged
needs: humanfor questions only you can answer. - A stale-claim rule that never sweeps up someone's unsaved edits.
How it was tested Built from a real two-day mobile app build (14 logged tasks), plus a dry run with two parallel agents fighting over one file. Every file ended with exactly one writer, and both agents logged what they did. The dry run produced the tie-break, UTC-with-seconds and commit-only-your-own-files rules.
Limits, stated up front
- It is an agreement between agents, not a file lock. An agent that ignores the rule can still edit anything.
- Tested in Claude Code. Codex and Cursor follow the same file through an AGENTS.md line, but I have not tested them.
- It does not make agents smarter; it makes their handoffs visible.
Part of the Agent Team Kit (free): run agents without chaos.
- Multi-Agent Ledger: claims, task log and handoffs in one LEDGER.md.
- Goal Loop: one agent working through a checklist until it's done or blocked on you.
Made by Armitage Beckett. MIT licensed.
Advisory only, not a file lock. Tested in Claude Code only.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 7 days ago
- Free with an Agensi account
Requires