Skip to main content
The recorder skill teaches an agent how to start a recording, capture the intended browser behavior, stop and upload it, and verify the result. It works alongside the CLI; it does not replace CLI installation or account authorization.

​
Install the CLI first

Complete CLI setup, including FFmpeg and browser requirements. Sign in with captures:record if the agent should upload recordings.

​
Download the skill

Download the Tekma recorder skill archive. Extract it and place the tekma-recorder directory in your agent’s skills directory. Common personal locations are ~/.codex/skills/tekma-recorder and ~/.claude/skills/tekma-recorder. Follow your client’s current skill-loading instructions and reload it after installation. The final directory should contain SKILL.md and its supporting references. Avoid an extra nested directory such as tekma-recorder/tekma-recorder/SKILL.md.

​
Ask for a recording

For example:
Use the Tekma recorder skill to record a short walkthrough of the example website. Upload it to my authorized workspace, verify that playback works, and return the share link.
For a bug fix, state the behavior the recording should prove. The agent should capture the actual interaction rather than substitute a static screenshot.

​
Permissions and privacy

The skill does not grant account access. Complete browser authorization yourself when required. It should never require you to paste an access token into a chat. The CLI’s recording limitations still apply, including the absence of element-level video redaction. Use synthetic data for demonstrations and inspect the resulting video.

​
Update

Install the latest CLI package and replace the skill directory with the current archive, then reload the agent. Keep personal modifications separately so you do not overwrite them accidentally.