Guides

AI agents write real code. These guides explain how to see what they changed, what to check before you merge, and how to keep a record you can stand behind.

• See exactly what Claude Code changed in your repo

  Four ways to review what Claude Code did in a session: git diff, the transcript, hooks and a per-session receipt, and what each one leaves out.

• How to audit the changes Cursor’s AI makes

  Cursor’s agent can edit many files at once. How to keep a record of the prompt, the files and the risks, and check it before you push.

• Record what Antigravity (Gemini) changed

  Google’s Antigravity agent editor has no prompt hook. How AgentReceipt still records the prompt, the files changed and the risk flags for each turn.

• What an AI coding agent audit trail should record

  A short checklist of what to record when AI agents write code in your company: the prompt, the diff, the flags, who asked, and how long to keep it.

• An AI code review checklist: 7 risks to check

  Seven things to check when an AI agent wrote the code: secrets, weakened tests, disabled checks, new dependencies, sensitive paths, scope drift, size.

• How to stop an AI agent committing secrets

  AI agents can copy keys and passwords into code and prompts. Practical steps to keep secrets out of your repository, and what to do if one gets in.

• How to catch an AI agent weakening your tests

  When an AI agent cannot make a test pass, it can change the test instead of the code. The signs to look for in a diff, and how to catch it automatically.

• What is AgentReceipt?

  AgentReceipt keeps a receipt of every AI coding session: what the agent was asked, what it changed and what looked risky. What it is and who it is for.

• How to review an AI-generated pull request

  A step-by-step way to review a pull request an AI agent helped write: start from the prompt, check the scope, then the risky parts, then the code.

• New dependencies added by AI agents: what to check

  An AI agent can add a package without asking. How to spot new dependencies in a diff, and the questions to ask before you accept each one.

• Scope drift: when an AI agent changes more than you asked

  Scope drift is when an AI agent edits files unrelated to your request. Why it happens, why it matters and how to catch it in review.

• How to review a very large AI diff

  Large diffs hide problems. Practical ways to split, order and review a big change made by an AI coding agent without missing the risky parts.

• Protecting sensitive code from AI agent mistakes

  Authentication, payments and database migrations deserve extra care when an AI agent edits them. How to mark them, review them and watch them.

• When an AI agent disables a lint or CI check

  Turning off a failing check is the quickest way to get a green build. How to spot disabled lint, type and CI checks in AI-written changes.

• Claude Code hooks explained, with an audit example

  Hooks let you run your own command when something happens in a Claude Code session. What they are, why they suit auditing, and how AgentReceipt uses them.

• Auditing Claude Code, Cursor and Antigravity compared

  How the three popular AI coding agents differ when you want a record of what they did, and how one tool covers all of them with the same receipt.

• A git pre-push hook that checks AI-written changes

  A pre-push hook runs a check on your machine before code leaves it. How to set one up so AI agent changes are checked before every push.

• Attaching AI session receipts to commits with git notes

  Git notes let you attach extra information to a commit without changing it. How AgentReceipt stores each AI session receipt as a note on the commit.

• Keeping secrets out of AI prompts and logs

  Prompts get pasted with error messages, and error messages contain keys. How to stop secrets ending up in prompt logs and audit records.

• An AI coding policy for teams: what to include

  A short, practical outline for a team policy on AI coding agents: where they may be used, what needs review, what must be recorded and who decides.

• Manager and employee views for AI coding audits

  Who should see what when AI agents write code: a manager sees every employee’s receipts, and each employee sees only their own.

• How long should you keep AI coding session logs?

  Keeping everything forever costs money and adds risk. How to choose a retention period for AI coding records, and what to do when it ends.

• A pull request template for AI-assisted changes

  Add a few lines to your pull request template so reviewers know when an AI agent helped, what it was asked, and where to look first.

• Prompt logging and privacy for AI coding tools

  Logging prompts helps review but raises privacy questions. What to log, who should see it, how to redact it and when to delete it.

• Write prompts that lead to reviewable AI changes

  A clear, narrow prompt gives a small, reviewable change. How to write prompts that keep AI agents inside the task you meant.

• A five-minute check of AI changes before you push

  A short routine to run before pushing code an AI agent helped write: diff, flags, tests, dependencies and secrets, in about five minutes.