- Home
- Skills
- Technical Documentation
- Release Notes
Release Notes
Stop shipping auto-generated commit logs that your users ignore.
$5
Works with the AI Agents you already use
Release Notes
Example session with this skill installed
Write the release notes for v2.4.0. The product's repository is in repo/ and an export of its issue tracker is in issues/.
- Read your context and instructions
- Compiled the release notes
A real, unedited run of this skill on a seeded test repository: Beacon is a fictional monitoring product with an 18-commit release and an issue-tracker export.
Range confirmed: v2.3.0..v2.4.0. That is 18 commits from 2026-09-08 to 2026-10-14, and the v2.4.0 tag is dated
14 October 2026. Two commits on main after the tag aren't in this release: SMS alerts (#270) and the status-page favicon fix (#271). They belong in 2.5.0.
Tier: Major. The diff is small, but one change breaks API integrations: the default page size drops from 500 to 50. There's also an API deprecation with a removal date and a security fix for SAML SSO.
Beacon 2.4.0 — 14 October 2026
Web app · Public status pages · REST API (v1)
You can now schedule maintenance windows that mute alerts and show "Under maintenance" on your status page. This release also fixes a SAML single sign-on security issue, gives uptime reports more accurate figures and makes check-result exports faster.
If you call the REST API, see
Action required: list endpoints now return 50 items per page by default, and /v1/webhooks is retired on 31 March 2027.
⚠️ Action required
List endpoints now return 50 items per page by default (was 500). This applies to GET /v1/monitors, GET /v1/checks and GET /v1/incidents, starting with this release. If your integration reads a whole list in one request without passing limit, it now receives only the first 50 items. You can fix this in one of two ways:
- follow
next_cursoruntil it's empty (recommended), or - pass
limitexplicitly. The maximum is still 500.
Large lists were timing out at the old default, so this change keeps those requests reliable. API reference: pagination →
Deprecated: /v1/webhooks. It will be removed on 31 March 2027. It works until then, and every response now includes Deprecation and Sunset headers so you can detect it automatically. Move to
/v2/hooks (available since 2.2.0). It delivers the same events and adds signed payloads and automatic retries. Webhooks v2 migration guide →
🔒 Security
SAML single sign-on: fixed a sign-in vulnerability (CVE-2026-41207). In workspaces using SAML SSO, a crafted sign-in response could have let an attacker sign in as another member of the same workspace. Beacon now rejects these responses. We found no sign of this being exploited. You don't need to do anything. The fix is already live for every workspace.
New
Maintenance windows: plan downtime without paging your on-call team or alarming your customers. While a window is open, Beacon mutes alerts for that monitor, and your public status page shows it as
Under maintenance instead of
Down. A window can be one-off or repeat every week. To set one up, open a monitor and go to
Settings → Schedule maintenance. Maintenance windows are turned on for every workspace.
Improved
- Incident timelines open faster in large workspaces with a lot of monitors and history.
Check-result CSV exports are faster. In our tests, a 90-day export of 1.2 million check results went from about 48 seconds to about 11 seconds.
Fixed
- Uptime figures missed outages that started on the last day of a month. For example, an outage on 30 September didn't count toward September, so the status page and monthly report could still show 100.00%. These outages are now counted.
- The weekly uptime report email never arrived for workspaces set up through single sign-on provisioning (e.g. Okta). It's now delivered every Monday. If your workspace has no timezone set, the report uses UTC dates.
Need help? Documentation · API reference · Support · Community forum
Notes for the reviewer
The release notes for 2.4.0 are ready. They contain:
- 1 feature
- 2 improvements
- 2 fixes
- 1 security fix
- 1 breaking change and 1 deprecation, both under Action required
- 10 commits dropped because users can't see their effect
How each commit was handled
| Commit | Bucket | Why |
|---|---|---|
| #251 maintenance windows | New | On for every workspace (the flag is only a kill switch). The flag isn't mentioned in the notes |
| #253 timeline index | Improved | Closes #309 |
| #254 digest null timezone | Fixed | Closes #318. Written from the issue's symptom ("weekly report never arrives") |
#255 /v1/webhooks deprecation | Action required | Date and replacement come from the PR |
| #256 CSV streaming | Improved | |
| #257 SLA month window | Fixed | Closes #322 |
#259 DEFAULT_PAGE_SIZE 500→50 | Action required | Labelled chore but it's a breaking API change. The commit body confirms that callers who don't pass limit now get only the first 50 items |
| #261 saml-verify 3.1.4 | Security | One of the skill's exceptions to dropping dependency bumps. It patches a CVE in a code path every SAML sign-in uses. The notes publish what the fix does, not which dependency changed |
| #252 + #265 Slack thread replies, then its revert | Dropped | Shipped and rolled back within this release. #300 is still open, so the notes don't mention it |
| #264 refactor | Dropped | The PR says "no behaviour change" |
#263 test, #262 CI, #260 eslint bump, #267 eslint autofixes, #266 docs typo, #268 vitest bump, chore(release) | Dropped | No effect users can see |
Where the numbers come from
CSV export (48 s → 11 s): the PR #256 body. It's the median (p50) of 10 runs on staging against the largest workspace snapshot (1.2M check results, 90-day export). The notes keep those conditions with the number. The memory figures from the same PR (1.9 GB → 140 MB) are left out because users can't see them.
Incident timeline: no number. The PR #253 author says it "feels a lot snappier locally" but wasn't benchmarked on production-sized data, so the entry says "faster" without a figure. To add one, measure timeline load time (p50/p95) for a large workspace before and after migration 0042. Production monitoring data from either side of the 14 October deploy would also work.
Before publishing
1.
Add the maintenance-window screenshot. The image is a placeholder (SCREENSHOT-TODO). Crop it to the monitor's Settings tab and use made-up data.
2.
Check that the links resolve. The doc links come from the README and PR #255, but I couldn't open them from here. In particular, the pagination link points to the general API reference because I couldn't find a dedicated pagination page.
3.
SLA fix: I couldn't tell from the code whether September's uptime is recalculated for past months, so the entry promises the fix only going forward. If historical months are recalculated, say so. The customer in #322 will want to know.
Distribution
Breaking change and deprecation: these justify a dismissible, time-limited in-app banner and a segmented email to API users. Workspaces with no API calls don't need the email.
- Maintenance windows: these fit the "What's new" panel, plus an inline tooltip on Schedule maintenance.
- SAML fix: worth telling SSO workspace admins directly.
CHANGELOG.md: replace the raw commit-subject format with this text (2.3.0 used commit subjects) so the repo changelog matches the public one.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop shipping auto-generated commit logs that your users ignore. Developers often mistake a list of merged PRs for a changelog, resulting in noise like "bumped eslint" or "refactored auth handler" reaching the end user. This skill forces a strict translation from code changes into user-facing value, ensuring your release notes actually answer the one question users have: what changed for me?
What it does
- Automatic Classification sorts changes into features, improvements, fixes, or the "drop" bucket for internal-only churn.
- Outcome Translation converts technical commit subjects into active-voice sentences describing new capabilities.
- Tier Calibration adjusts the document depth based on whether the release is a patch, minor, or major version.
- Breaking Change Prioritization moves required actions to the top of the document regardless of diff size.
- Evidence Verification flags improvements lacking measured benchmarks to prevent fabricated performance claims.
How it works
- Gather the change set using
git logranges or PR titles from the repository history. - Classify every change into functional buckets while aggressively dropping internal refactors and CI updates.
- Translate surviving entries into user-facing prose, removing all internal class names and jargon.
- Assemble the notes in a specific hierarchy: Header, Summary, Breaking Changes, then Features.
Frameworks & tools
Works with any git-based repository. It uses standard git log conventions and is compatible with GitHub PR titles and issue trackers for context.
Why this beats prompting it yourself
Generic prompts often hallucinate excitement or include "internal dependency updates" that users don't care about. This skill uses a strict "could a user notice this?" filter and a non-negotiable rule against engineering jargon, ensuring high signal-to-noise.
Use cases
- Generating a "What's New" modal for a SaaS product launch.
- Writing a public-facing
CHANGELOG.mdfor an open-source library. - Preparing high-level updates for a mobile app store submission.
- Communicating breaking API changes to downstream developers.
Known limitations
Does not support non-git version control systems. Requires access to PR descriptions or linked issues to provide high-quality context for complex features.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Technical Documentation.
Requires git history access. Quality depends on PR titles or issue descriptions. Does not generate visual assets like GIFs.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 14 days ago
- Software
Git