/next/organisation/ai-skills) is where you teach it: Skills are Markdown instructions it reads when the work calls for them, and Custom tools are scripts it can run for you in a locked-down sandbox.
This page covers what Ankra AI uses inside the platform. To teach your editor
or assistant (Claude Code, Cursor, Copilot and others) Ankra’s conventions, install the
Agent Skills instead.
Skills
Skills teach Ankra AI how your team works: runbooks, add-on configuration templates, CI/CD pipeline conventions. Ankra ships its own, and people who manage Ankra AI can replace any of them with their own copy or add skills of their own. Find a skill searches the list, and Kind filters it to Core, CI/CD or Add-on skills.How skills resolve
Skills come in two tiers:- Ankra’s skills - built in and maintained by Ankra. You can read them but not change them.
- Your organisation’s skills - added by your organisation.
Add a skill
1
Open the editor
Go to Ankra AI → Skills and tools and press Add skill. To start from one of Ankra’s skills instead, open its options menu and choose Replace with your own copy….
2
Fill in the fields
- Name - lowercase letters, numbers, dashes and underscores. A skill’s name does not change once it is saved.
- Add-on charts (optional) - the charts this skill is about, separated by commas. Ankra AI reads it when one is involved.
- Description - one sentence. Ankra AI reads it to decide when the skill applies.
- What the skill teaches - the instructions, in Markdown, up to 200,000 characters.
- Always in use: Ankra AI reads it in every conversation - switch it on to make Ankra AI read the skill every time.
3
Save
Press Add skill, or Save your copy for a replacement. The skill is listed under Skills.
Who can change skills
Everyone who can open the page can read the skills. Adding, changing and deleting them needs a role that manages Ankra AI, such as owner or admin. Ankra’s own skills cannot be changed or deleted - only replaced with your own copy.Custom tools
Custom tools extend what Ankra AI can do with capabilities specific to your organisation: a security scan, an internal API call, a report generator. Each tool is a script with an input schema; once it is turned on, Ankra AI can call it like any built-in tool, ascustom__<name>.
The sandbox model
Each run happens in a locked-down sandbox on a cluster:- a signed runner image,
- no service account, so no access to the Kubernetes API,
- no network unless you allow it,
- a read-only file system,
- a time limit you set.
TOOL_INPUT environment variable. Whatever the script prints is returned to Ankra AI.
Add a tool
1
Open the editor
Go to Ankra AI → Skills and tools, scroll to Custom tools and press Add tool. The Add a custom tool dialog opens.
2
Define the interface
Give it a Name in snake_case and a Description of what it does and when Ankra AI should use it - the description is what Ankra AI reads when deciding to call it, so make it concrete. In Input schema, give a JSON Schema object: what Ankra AI must give the tool.
3
Write the script
Choose where it Runs in (Python 3.12, Node.js 20, Shell or Trivy), set the Time limit, in seconds, and write the Script.
4
Set the guardrails
- Risk - Low risk, Medium risk or High risk.
- Where it may run - In a sandbox on any cluster, or Only on the clusters I name with Clusters it may run on.
- Secret files - secrets the script reads as files under
/run/agent-secrets/. Never write a credential into the script. - May reach the internet over HTTPS - off by default.
- Runs without asking: nobody approves each run - off by default. Leave it off unless you would approve every run.
- Turned on: Ankra AI may use it.
5
Save
Press Add tool. The tool is listed under Custom tools, with its risk and where it runs.