How to make videos with GitHub Copilot CLI

1. Keep the episode in a repo: scripts/ for narration and narration/ for recordings. Start copilot in that folder with read-only Cliptude tools pre-approved: copilot --allow-tool='cliptude(list_templates), cliptude(estimate_credits), cliptude(get_job)'.
2. If you recorded the narration yourself, ask Copilot to upload narration/episode4.mp3: it calls start_voiceover_upload with the file name and size, PUTs each part to the signed URLs with a shell command, and calls finish_voiceover_upload for a voiceover_url. If not, ask it to write the narration with /cliptude-script.
3. Ask for estimate_credits, then create_job with your template, the voiceover_url or script, and a thumbnail reference video. Copilot shows Cliptude's review; you confirm, and it sends create_job again with the review_token.
4. Type /every 10m check Cliptude job <id> with get_job and tell me when it is completed. Copilot runs the check on that interval while the session stays open; type /every with no arguments to remove it.
5. When the video is done, ask for open_editor_project and view_scenes from the first scene to the last, fix any wrong shot with run_agent, and render_project.
6. For a draft every week without the session open, run copilot -p from cron with only list_templates and estimate_credits allowed, so it writes and prices the script but cannot start a paid job.

Source: GitHub Docs: Schedule prompts in Copilot CLI

Copilot CLI loads agent skills, folders with a SKILL.md file. It picks a skill on its own when your request matches the skill's description, or you name it with a leading slash, such as "Use the /cliptude-script skill to write the narration."

It looks in a repository's .github/skills/, .agents/skills/ and .claude/skills/, then in ~/.copilot/skills/ and ~/.agents/skills/, then in plugin folders and any folders listed in COPILOT_SKILLS_DIRS. If two skills share a name, the first one found wins.

The skills CLI's github-copilot target writes to .agents/skills/ for a project and ~/.copilot/skills/ with -g, both folders Copilot CLI reads. Copilot also has its own commands: copilot skill add takes a file, URL or folder, and inside a session /skills add, /skills list, /skills info and /skills reload manage them. Skills are switched on and off with copilot skill enable and copilot skill disable.

Plugins are a separate layer: /plugin install takes a marketplace spec, GitHub repository, git URL or local path. Cliptude does not publish a Copilot CLI plugin, so install the skills and add the MCP server as above.

npx skills add cliptude/cliptude -a github-copilot -g

git clone --depth 1 https://github.com/cliptude/cliptude /tmp/cliptude-repo
copilot skill add /tmp/cliptude-repo/mcp_server/skills/cliptude
copilot skill add /tmp/cliptude-repo/mcp_server/skills/cliptude-script
copilot skill add /tmp/cliptude-repo/mcp_server/skills/cliptude-plan

Source: GitHub Docs: Adding agent skills for GitHub Copilot CLI

Both work with the same Cliptude server and skills, and both can upload a voiceover from your disk because they run shell commands. Copilot CLI is the low-cost start if you already have a GitHub Copilot plan, and it reads skills from .claude/skills, so a repo set up for Claude Code works as is. Claude Code is the better fit if you want a weekly video run in the cloud with your computer off; Copilot's /every schedules stop when you close the session.

Source: GitHub Docs: Schedule prompts in Copilot CLI

Copilot CLI needs an active Copilot plan, and the Free plan includes it with limited chat and agent usage. Paid plans for individuals are Pro at $10 a month, Pro+ at $39 and Max at $100, each with GitHub AI credits; extra usage costs $0.01 a credit under a budget you set. Copilot CLI and agent mode draw on those AI credits, at a rate that depends on the model and the task.

Copilot's credits pay for the agent's writing and tool calls. The video is charged separately to your Cliptude credits, on a plan from $19; Cliptude has no free plan.

Source: GitHub Copilot plans

GitHub Copilot CLI is the command-line interface for GitHub Copilot: you run copilot in a terminal and work with Copilot's agent there, interactively or with -p for a single prompt. It reads and edits files, runs commands with your approval, and can be extended with MCP servers, skills, custom agents, hooks and plugins. It needs a Copilot plan, and the Free plan includes the CLI.

GitHub Copilot CLI or Copilot in VS Code?

Both are Copilot's agent and both take MCP servers. The CLI keeps its servers in ~/.copilot/mcp-config.json and suits scripts, scheduled prompts and voiceover uploads from the shell; VS Code adds Cliptude with MCP: Add Server.