# Status Cadence

> Check one API incident's public status-page updates against the skill's incident-communication practice, then
> rewrite the timeline and draft the public postmortem. Paste the updates (time | status | message, or the status page's incident view as it copies), the components
> you flagged and the time the monitoring alert fired; the browser checks the four-stage lifecycle, holds every
> next-update promise to the next post, measures detection-to-disclosure lag and whether a third party reported
> first, and flags missing time zones, codename titles, vendor deflection, named individuals, speculation and
> severity that is not breadth of impact. Built on the SkillSafe skill @samber/api-status-communication, from
> samber/developer-platform-skills (MIT), by Samuel Berthe.

https://status-cadence.skillsafe.ai/

## What is free
- The lifecycle: investigating, identified, monitoring and resolved, each timestamped; updates after Resolved; a newest-first list sorted before checking.
- The cadence: every gap between posts, every next-update promise (relative, "in 30 minutes", or absolute, "by 10:15 UTC") held to the next post with five minutes of grace, and the longest silence.
- Disclosure: the first public post against the monitoring alert (ten minutes or more fails - a human gate is still in the loop) and against the first third-party report.
- The wording: a plain-language title with no ticket number or codename, a zone on every time, no vendor blamed as the cause, no person named, no speculation, no blanket "the API is down" the components contradict, the impact acknowledged.
- The components: severity as breadth (a subset is partial, everyone is major), consumer-recognizable names, the top-level status as the worst component.
- The practice: the channel ladder (page with email and RSS, webhook push, SMS push, out-of-band redundancy) ranked from the interview; the postmortem depth (resolved note, status-page summary, engineering-blog postmortem); the outage in SLA minutes at 99.9%, 99.95% and 99.99%.
- Exports: a cadence table (Markdown), the timeline (.csv), the checks (.csv) and everything (.json).

## The metered lanes
Two tasks on gpt-terra, metered per run, routed by `task`. Input string fields: `task` (rewrite or postmortem), `incident`, `timeline`, `carry` (postmortem only: the rewritten timeline) and `facts`.
- `rewrite` replies with `summary`, `title` and `updates` (one per posted update: `n`, `status`, `text`) and `caveats`.
- `postmortem` replies with `summary`, `rung`, `title`, `impact`, `timeline` (`time`, `event`), `cause`, `worse`, `fix`, `prevention` (`action`, `owner`) and `caveats`.
The page holds every reply to its own facts: one rewrite per posted update with its status, every open update ending with the next-update time the real timeline kept, every time zoned, no vendor blamed, no person named; the postmortem on the ranked rung, the impact window in minutes, only times the input gives, and an owner on every prevention item.

## Links
- API tutorial: https://status-cadence.skillsafe.ai/api.html
- Source skill: https://skillsafe.ai/skill/@samber/api-status-communication