how it works
What is a Skill?
A skill is a defined operational workflow that Hyground executes on your behalf. It knows which data sources to query, what evidence to collect, what actions to take, and how to present findings.
Every team member executes the same workflow with the same rigour, whether it is 3pm or 3am.
You create skills yourself: describe one in a chat, start from a blueprint, or write it in the editor. See how our team turned a regression test into a skill.
Agents autonomously load skills on demand
Every skill's description sits in the agent's system prompt, but the full body is fetched only when the agent decides to use it. The agent calls `skill(name, reason)`. The description tells it what the skill does and when it applies. The actual procedure is pulled in just-in-time. Context stays lean even as the skill library grows into the hundreds, and each load is recorded with the agent's stated reason.
Skill blueprints
Start from a ready-made use case instead of a blank page. Pick a blueprint and click Create in Hyground. The agent explores your environment, checks the blueprint’s assumptions against what it finds and asks a few questions about your setup, then writes the skill with you and offers a test run.
01
Takes a crash, stack trace or exception from logs and locates the responsible code in the repository, including the recent changes that touched it.
02
Turns a diagnosed problem or a routine ticket into a proposed code or configuration change, opened as a pull or merge request for review.
03
Performs a first analysis of new support tickets so first-level support can resolve or redirect known cases before they reach developers.
04
Validates a pending GitOps or manifest change before it rolls out: image tags exist, routes and hosts resolve, and references are consistent.
05
Checks each application’s repositories against your internal engineering and security rules and rates every rule with a clear to-do list.
06
Assembles the timeline, impact, root cause and follow-up actions of an incident into a structured report ready for review.
Custom skills with your operational specifics
Built-in skills cover the common cases. Custom skills encode your organisation’s specifics, your architecture, your runbooks, your definitions of healthy, your remediation steps.
01
Plain text or structured format, whichever works for your team
02
Automatically shared across your team when saved
03
Bundle a Python or Bash script alongside skill.md; the agent runs it in a sandboxed workspace with a 30-second timeout
04
Skills can reference any API or MCP server you have configured
Example
Check payment-service pod health in prod
Verify Stripe webhook endpoint is reachable
Confirm database connection pool utilisation < 70%
Check for any open P1 incidents on dependent services
Confirm last deployment succeeded without rollback
Operational Knowledge That Survives Attrition
Every new skill added to Hyground can be shared and is executable by every person on your team. Over time, the platform accumulates the operational expertise of your entire organisation, independent of individual people.
When a senior engineer leaves
When a new engineer joins
