Who it is forA developer or auditor who needs a reproducible, checkable review of one target for a fixed time window.
The problemA quick chat audit cannot be reproduced or handed off, so reviews need explicit inputs and a checked report.
The outcomeA validated audit scaffold and a review-check result for a final report written to an approved non-home path.
What you need first
- Git
- Codex with local plugin support
- Python 3 for the requirement-ledger CLI, installed with uv or pip
- A local project folder to select as the target
Steps
- From the checkout, run
uv tool install .. - Run
requirement-ledger --versionand confirm the release you installed. - Run
requirement-ledger review-init --mode audit --target project:example --start 2026-08-01T08:00:00+08:00 --end 2026-08-02T08:00:00+08:00 --timezone Asia/Shanghai --output /approved/review/final-report.md, replacing the example values with your own approved values. (inferred, check the README) - Run
requirement-ledger review-check /approved/review/final-report.mdand read the result before treating the report as checked.
Prompt for your agent
Using the requirement-ledger CLI from the gpt-6-astra-skill repo, create an audit scaffold for project:<target> over the window <start> to <end> in timezone <timezone>, written to the approved non-home path <output>. Return the report path and the review-check output. Verify the review-check output before calling the report checked.
Caveat. The CLI is the compatibility core, and passing it does not by itself prove that every product or public-release gate has passed.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for adand-91/gpt-6-astra-skill and "Run an evidence-bound review with the requirement-ledger CLI"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer