Prompt change checklist for agent teams

Before a prompt reaches production

[ ] Diff the new version against the last known-good one — see the semantic delta, not just the text.
[ ] Regression check — confirm no guardrail or instruction present in A was lost in B.
[ ] Gate the merge — a failed regression check does not ship (mirrors Art. 15 robustness control).
[ ] Log the change — version, diff, regression result, timestamp (Art. 12).
[ ] Export the history so the change is reproducible for audit.

After every change

[ ] Re-run the regression on the next edit — drift is the default, not the exception.
[ ] Re-check that no new high-risk instruction is unguarded.
[ ] Re-validate the obligation map (Art. 9 / 12 / 15) with current effective dates in mind.

What the checklist is — and is not

The checklist is a control inventory for prompt change. It tells you which guards held and which a change dropped. It is not a pass/fail certification, and ticking every box does not make the system "compliant." Conformity is the deployer's responsibility across the full system and should be confirmed with qualified counsel.

Using AgentPrompt in the checklist

Paste version A and B, choose diff + regression, and AgentPrompt returns the behavioral delta and any lost guardrail. Keep the history, roll back a bad edit, and export versions for audit — then re-run on every release.

Authoritative references

EU AI Act (Reg. 2024/1689), Art. 9 / 12 / 15: https://eur-lex.europa.eu/eli/reg/2024/1689/oj
AI Act explorer: https://artificialintelligenceact.eu/
OWASP LLM Top 10 (2025): https://owasp.org/www-project-top-10-for-large-language-model-applications/

FAQ

What belongs in a prompt change checklist?

Diff the new version against the last good one, run a regression check for lost guardrails, gate the merge on a clean check, log the change (Art. 12), and export the history for audit.

Should a failed regression check block the change?

Yes — for the same reason a failing unit test blocks code. For high-risk uses this gate is also your evidence for Art. 15 robustness.

Does a completed checklist certify compliance?

No. It is a control inventory for prompt change — evidence of which guards held. Conformity is the deployer's responsibility across the full system and must be confirmed with qualified counsel.

Which EU AI Act articles map to prompt governance?

Art. 9 (risk management), Art. 12 (logging), Art. 15 (robustness). Versioned, regression-checked prompts help evidence all three.

Is AgentPrompt a legal opinion?

No. AgentPrompt helps you version, diff, and regression-test prompts. It is decision-support, not a legal opinion, a certification, or a guarantee of conformity.

## Disclaimer

Disclaimer (decision-support, not legal advice): AgentPrompt is a tool that helps teams version, diff, and regression-test agent prompts. It is not a legal opinion, a compliance certificate, or a guarantee that an agent is safe or compliant. EU AI Act conformity depends on the full deployed system and is the responsibility of the deploying organization. Verify current obligations and effective dates with qualified counsel before relying on any timeline.

Try AgentPrompt to put a regression-gated checklist on every prompt change — try AgentPrompt.