What you’ll get: Skills let your team encode institutional knowledge, “when you see X, investigate like this”, and have ewake apply it automatically, without you repeating the instructions every time.
This page is about the playbooks ewake follows during investigations. For the skills a coding agent runs to set up ewake in your repository, see Setup skills for coding agents.
What it does
Skills are playbooks written by your team that guide ewake’s behaviour in specific situations. When an investigation starts, ewake is given a catalog of your active skills, one line each: the title and the description, and nothing else. As the investigation reaches a subject one of them covers, ewake loads that skill in full and weighs it against the evidence it is gathering. That is why the description carries so much weight: it is what ewake reads when deciding whether a skill is worth opening. The result: ewake responds the way your team would, based on the guidance you’ve already written, more consistent answers, less repetition.Anatomy of a skill
Each skill has three fields:Statuses
A skill moves through the following statuses:Versioning
Every time you save changes to a skill, a new version is created, and that version immediately becomes the current one — the one ewake uses. Saving is publishing; there is no separate review step. You can view the full version history and put an earlier version back in service with Activate this version. Doing so promotes it to current without deleting anything that came after it.Skills ewake writes itself
Ewake writes skills through two paths, and they do not behave the same way.From material you point it at
Show ewake a document — a Notion page of runbooks, a written post-mortem, a Slack thread — and ask it to learn from it, and it writes a skill from that text, or refines the existing skill that already covers the same ground. A new skill written this way arrives as a draft and does nothing until someone activates it in the dashboard. A refinement of a skill you already have behaves like any other save: the new version becomes the current one straight away. If that skill is active, ewake uses the refined body from the next investigation on; if it is in draft, disabled or archived, the version is updated but nothing is served until you activate the skill.From feedback on an investigation
When someone rates an investigation 👍 or 👎 in Slack, ewake re-reads that run against the feedback and may sharpen an existing skill or write a new one. See Giving Feedback. Both the edits and the new skills go live immediately — that is the point of the loop: a correction you give today applies to tomorrow’s alert without anyone having to promote it. Either way, the new version carries a per-field explanation of what changed and why — and, for the feedback path, which feedback and which investigation it came from. Both are visible in the skill’s version history, and any change can be undone by activating an earlier version, disabling the skill, or archiving it.How to create a skill
1
Go to Skills
In your ewake dashboard, navigate to the Skills section.
2
Create a new skill
Click New skill.
3
Write the title
Choose a title that clearly describes the situation this skill covers.Examples:
Database connection pool exhaustionHigh 5xx rate on payments-apiPost-deploy latency spike triage
4
Write the description
Describe when this skill applies. This is what ewake uses to match the skill to a mission.You can click Suggest to have ewake generate a description from the title and body.Example:
5
Write the body
Write the guidance ewake should follow. Treat it as a runbook: concrete steps, specific things to check, how to interpret results.Example:
6
Save and activate
Tips for writing effective skills
Be specific. A skill scoped to “database connection pool exhaustion on checkout-api” is more useful than one titled “database issues”. Specific skills get retrieved more precisely and contain more actionable guidance. Write the body as instructions to the agent. Use numbered steps, concrete thresholds, and named services. Avoid vague directives like “investigate carefully”. Use the description as a trigger condition. The description should answer: in what situation should ewake apply this? Think of it as theif clause.
Start in draft, activate when ready. There is no cost to keeping a skill in draft while you refine it. Activate only when you are confident the guidance is correct.
Need help? Contact support@ewake.ai.