PUBLIC BETA / V0.3.3

One real task. Fifteen minutes.

Use Agent Ledger on a small change you would already give to Codex or Claude Code. Inspect the evidence, then tell us whether the loop made the work easier to trust.

Begin the test Send feedback

15 minTotal target time

1 taskSmall and reversible

0 uploadsNo run data is sent

00-02Install

Set up v0.3.3 in a safe repository.

Requirements: Git, Node 20+, and Codex or Claude Code. Start from a clean branch when practical.

npx skills add sprintagency-it/agent-ledger --skill agent-ledger --yes --copy

The CLI detects your agent; add --agent codex or --agent claude-code if needed. For both agents at once, use npx --allow-git=all --yes github:sprintagency-it/agent-ledger#v0.3.3 setup --project .. Set DISABLE_TELEMETRY=1 to opt out of the skills CLI's anonymous install telemetry.

02-10Run

Give the agent one concrete outcome.

Use a real task, but keep the first run away from deployments, auth, secrets, billing, destructive operations, and sensitive client data.

# Codex
Use $agent-ledger to make this small change: <task>. Stay within <paths> and run <check> before handoff.

# Claude Code
/agent-ledger Make this small change: <task>. Stay within <paths> and run <check> before handoff.

10-13Inspect

Check whether the record matches reality.

executive-summary.mdCan you understand the outcome without reopening the chat?

review.jsonAre the status, evidence, and findings structured for another agent?

fix-brief.mdWere important signals classified before correction?

replay.htmlCan a human see what happened and what remains unresolved?

13-15Report

Tell us what was useful and what slowed you down.

The form asks about completion, useful catches, false positives, added friction, and whether you would use Agent Ledger again. A text-only response is enough.

Submit beta feedback Compare example reports

PRIVACY BOUNDARY

Your run stays local.

Agent Ledger has no hosted account and sends no run data to the maintainer. Raw diffs, snapshots, transcripts, and command logs remain under the ignored .agent-ledger/ directory.

The smaller share/ bundle uses best-effort redaction. Inspect every file before attaching anything; uploads are optional.

COMPLETION RULE

An install is not a beta result.

A completed run means the task finished, the tester inspected the summary and replay, and structured feedback was submitted. That is the evidence we will use to decide what to build next.

Start now Read full docs