Solve competition mathematics problems with rigorous proofs. Activates when asked to 'solve this IMO problem', 'prove this olympiad inequality', or 'find all functions satisfying'.

Invoke this skill proactively whenever a command needs Snowflake data access.

Take an application from an idea or an existing repository to a running cloud deployment. Use when the user asks to deploy an app, plan its infrastructure or estimate what it will cost to run.

Formats code across multiple languages

Exercise invalid OpenAI skill metadata.

Skill that handles database backup operations

Use when a berth allocation change needs auditing before the timetable is published

Use when a sailing has no berth assigned, to find which allocation constraint rejected it and what would satisfy that constraint.

Draft a changelog entry from merged pull requests when the user prepares a release or asks what changed since the last tag.

Connect a notes workspace when the user asks to link an account or check access to shared meeting notes.

Exercise legacy recursive skill discovery in a dual Agent and Claude package.

Automate deployment steps and environment bootstrapping for the service.

Draft release notes from the merged pull requests since the last tag. Use when preparing a release or summarizing recent changes.

Audits the pipeline configuration for duplicate stage definitions.

This skill lives outside the plugin root and must never be loaded.

Runs an internal workflow.

Runs an internal workflow.

Legacy review checklist that Cursor ignores because the manifest declares capabilities/. Use when another agent asks for the older checklist format.

Summarize open incidents from the pager rotation and produce a morning handoff report. Use when the user asks for an incident summary, a handoff report, or the current pager status. Includes links to each incident, the owning team, current severity, and the latest status update for every open incident.

Investigate a production incident from start to finish. Use when the user reports an outage, a spike in error rates, or asks for an incident investigation. Collects the recent deploy history, correlates it with the alert timeline, and pulls the relevant service logs for the affected window. Summarizes the most likely root causes with supporting evidence and links to the dashboards used. Prepares a draft incident report with an impact assessment, a timeline of events, and the follow-up actions the team agreed on. Posts the report to the incident channel and files tracking issues for each follow-up action so nothing is lost after the incident is closed. Keeps a record of every command it ran so the investigation can be audited or replayed later. Escalates to the on-call engineer when the evidence points at an ongoing problem rather than a resolved one. Works across the API, worker, and frontend services and understands the standard deployment topology. Never restarts services or rolls back deploys on its own; it always proposes the action and waits for confirmation. Understands maintenance windows and will not page anyone for alerts that fired inside one. Falls back to read-only analysis when it lacks credentials for a service. Reports partial findings rather than failing outright when a data source is unavailable.