Skip to content
Templates/Maintenance
MaintenanceLow risk

Release Notes Loop

Draft a changelog or release notes from merged PRs and commits, grouped and human-reviewed before publishing.

What this Loop Engineering template does

Draft release notes from the merged PRs and commits since the last release, grouped by type, for a human to review and publish.

Work only from the actual git history and merged PRs since the last tag. Do not invent entries. Group by Added / Changed / Fixed.

When to use it

Drafting changelogs from merged PRs
Grouping changes by type
Pre-release summaries

When not to use it

Publishing or tagging releases without human approval
Inventing changes not in the history

Validation checks

validation
Every entry maps to a real commit or PR
Entries are grouped by type (Added / Changed / Fixed)
Breaking changes are called out
No entries are invented

Boundaries & stop rule

!Do not publish or tag a release without human approval
!Do not invent changes that are not in the history
!Do not omit breaking changes
!Do not edit code
Stop rule — Stop when the draft covers all changes since the last release and is ready for human review. If a change cannot be traced to a commit or PR, flag it as unverified instead of including it.

Copy the loop prompt

claude-goal.txt
/goal Draft release notes from the merged PRs and commits since the last release, grouped by type, for a human to review and publish.
 
Task type: Custom
Target tool: Claude Code
 
Work toward this goal until all validation checks pass or the stop rule is reached.
 
Design hypothesis:
This release notes loop can produce a safer result if the scope stays narrow, validation is explicit, and a checker can reject shortcut work.
 
Smallest useful run:
Run one bounded pass on the newest relevant signal before expanding scope or adding a schedule.
 
Loop cycle:
1. Discovery — Read the latest signal for this template before acting: CI output, issue detail, review comment, dataset report, or content brief.
2. Handoff — Hand the work to one agent in an isolated branch, worktree, or clearly scoped session. Keep final approval with a human.
3. Verification — Use an independent review pass to confirm the result, inspect the diff or artifact, and reject shortcut work.
4. Persistence — Save a short run note with the signal reviewed, actions taken, validation result, and next recommended step.
5. Scheduling — Run manually until the loop is reliable; only then consider a scheduled or event-triggered run.
 
Context:
Work only from the actual git history and merged PRs since the last tag. Do not invent entries. Group by Added / Changed / Fixed.
 
Validation:
Every entry maps to a real commit or PR
Entries are grouped by type (Added / Changed / Fixed)
Breaking changes are called out
No entries are invented
 
Validation evidence:
Record the original signal, checks run, final result, changed files or artifacts, and any checker rejection.
 
Independent checker:
Use an independent review pass to confirm the result, inspect the diff or artifact, and reject shortcut work.
 
Boundaries:
Do not publish or tag a release without human approval
Do not invent changes that are not in the history
Do not omit breaking changes
Do not edit code
 
Stop rule:
Stop when the draft covers all changes since the last release and is ready for human review.
Maximum iterations: 2
 
Budget:
Example only: stop before exceeding the agreed per-run token budget.
 
Human approval:
Required before merge, deploy, delete, purchase, or external communication.
 
Fallback:
If a change cannot be traced to a commit or PR, flag it as unverified instead of including it.
 
Loop Validation Log:
- Hypothesis: This release notes loop can produce a safer result if the scope stays narrow, validation is explicit, and a checker can reject shortcut work.
- Smallest useful run: Run one bounded pass on the newest relevant signal before expanding scope or adding a schedule.
- Expected evidence: Record the original signal, checks run, final result, changed files or artifacts, and any checker rejection.
- Actual evidence: [fill in after the run]
- Passed? [yes / no / partial]
- Feedback: After the run, note what the loop learned, what failed, and what should change before the next pass.
- Next step: [stop / adjust the loop / run the next pass]
 
Do not delete tests, bypass checks, or modify unrelated files just to satisfy the validation condition. If blocked, stop and summarize the blocker, attempted fixes, and recommended next action.

Failure modes to watch

Invented changelog entries
Missed breaking changes
Publishing without review
Ungrouped, unreadable notes
Worked scenario review · maintained by TianMingAI · 2026-07-21

Review receipt

This is a bounded review exercise for the template, not a claim about a production deployment.

Scenario

A release contains a new lab-notes section, route consolidations, and an internal advertising pause but no user-account changes.

Baseline evidence

The writer compares the release range, issue decisions, route map, user-visible pages, known limitations, and deployment status instead of relying on commit titles alone.

Validation result

The notes distinguish additions, changed URLs, and operational changes, include migration implications, omit internal-only noise, and do not claim deployment before the live check succeeds.

Shortcut rejected

Publishing a raw commit list or calling every internal refactor a feature is rejected because it obscures what users and site operators must know.

Human gate

The release owner confirms scope, wording, known issues, and actual deployment state before the notes become public.

Final decisionPass

Loop Engineering FAQ

Every entry must map to a real commit or PR, and inventing entries is a forbidden action. Anything that cannot be traced is flagged as unverified.