Availability: Skills are being rolled out. If you do not see Build a skill in Chat or Skills in Workshop, this feature is not yet available to your workspace.
A skill is a reusable set of instructions that helps Hatz perform a particular task. For example, you can build a weekly client status report skill that turns project notes and task counts into a short update with Progress, Risks, and Next steps—without sending the report for you. After it is validated, you can select it in Chat whenever you need that report.
Build the first draft
Open Chat and choose Build a skill.
Answer the guided questions: what the skill should do, your role, what it should help Hatz handle, when it should be used, what it should produce, and whether it should try a tool. Only the first answer is required; skip questions that do not apply. Use Other if a listed answer does not fit. Be specific about boundaries: for example, “Draft the report only; do not send messages.” If your report includes task counts, you can select Python Code Execution to calculate a completion percentage.
Review the Skill proposal. Open Content to read the instructions before uploading. Check that the skill does what you intended and does not include actions or data it should not use.
Refine it with Regenerate
If the draft needs work, enter a concrete change under Regenerate this draft with a prompt and choose Regenerate. For the report example: “Keep the update under 150 words. Use headings Progress, Risks, and Next steps. Do not send anything externally.” Review the replacement draft under Content, then choose Upload skill when you are satisfied.
Read the scan result
Uploading starts validation and a security scan. The version can show:
Pending: Checks are still running (the skill’s page shows this as Validating). Wait for the result before testing it.
Validated: The version passed validation and is available in Chat. Read its scan verdict before relying on it.
Blocked: The version failed bundle validation or the malware check and is not available. Fix the bundle and upload a new version.
A validated version also gets a scan verdict. The verdict is advisory feedback on the skill’s contents; it does not change whether the version is available in Chat:
No concerns identified: The scan found no concerns. Still review the instructions yourself.
Caution: Read the findings and decide what to revise. If the report says the deeper analysis did not complete, review the skill yourself before using it. If an attempted scan did not finish, do not treat it as a clean result.
Not safe: The report found serious concerns. The version can still be selected in Chat, so do not use it; remove the risky behavior and upload a new version.
Scan incomplete: The scan did not produce a reliable verdict. Do not treat it as a clean result; contact Support if the issue persists.
Expand each finding to see what was flagged and why. If the explanation makes sense, use Regenerate with specific feedback, review the new draft, then Upload new version. A blocked version never replaces an earlier validated one.
Example scan report; your findings will differ.
Regenerate after upload
Want to improve the uploaded skill? Start a new chat, then in the message box open the Skills and tools menu, choose Skills → Manage skills, then choose … → Edit next to the skill. Read Content, give Regenerate a specific change, and review the replacement draft. Regenerating does not change the uploaded version. Choose Upload new version to submit the replacement for its own scan. Until that new version validates, Chat continues to use the latest validated version, if one exists.
Test the skill in Chat
Start a new chat. Type / in the message box to open the Skills picker, choose your validated skill, and confirm its highlighted /<skill-name> mention appears in your message. Then ask for a sample result with enough context to complete it—for example, a client, project, reporting week, notes, and task counts. Check the output against the skill’s instructions. If it misses a requirement, start a new chat, open the Skills and tools menu in the message box, choose Skills → Manage skills, then choose … → Edit next to the skill. Regenerate with a precise change and upload the new version.
Already have a skill?
If you already wrote a skill outside Hatz, you can upload it instead of building it in Chat. In Workshop → Skills, choose Create → Upload skill and select the skill’s files or folder. Put SKILL.md at the root; it needs YAML frontmatter with a name and description, followed by instructions. The name can use only lowercase letters, numbers, and single hyphens, up to 64 characters; the description can be up to 1,024 characters. The bundle may contain up to 20 files and 20 MiB total; each file must be at most 1 MiB, and SKILL.md at most 256 KiB. Supported file types include Markdown, text, JSON, YAML, CSV, Python, shell, JavaScript, PDF, Word, and Excel files. Uploading a new version follows the same scan and validation flow.
If upload preparation fails, use the message shown to fix the bundle. Two common messages are “SKILL.md is missing — place it at the root of your skill folder.” and “Skill bundles are limited to 20 files.” If a version is blocked after upload, the upload dialog shows “This skill has security risks”. That version will not be used; fix the bundle and upload a new version.