<- all tokdocs

Full Send Skill for AI Agents: What the Nerd Snipe Clip Leaves Out

By , founder of Mavric and Apso

View on TikTok ->

Full Send is an agent skill, described on the Nerd Snipe podcast, that tells a coding agent to fix a problem, verify it, open a pull request, wait for CI and review bots, and merge the PR itself. The short clip makes the merge sound like it depends only on the agent feeling confident. The full episode adds an "8 out of 10" confidence threshold, a second-model review loop and a sibling skill that stops before merging, and GitHub's own branch rules decide whether the merge can happen at all.

What is the Full Send skill?

A TikTok from @nerdsnipepod opens with "I created a skill called Full Send." The speaker then lists the steps: fix, verify, "file a pull request, babysit the pull request until everything passes," wait for CI and "auto reviewers," do "one last verification pass," and "if your confidence is high, merge it and assume I will never see this again."

The clip comes from "Nerd Snipe with Theo and Ben," a weekly show hosted by Theo Browne ("CEO @ Ping.gg" on GitHub) and Ben Davis. The same cut went up on YouTube Shorts on October 13, 2026. Its parent is episode 25, "OpenAI's Dev Day Was a Disappointment", dated October 12, 2026 in the podcast feed per Castro.

The clip's exact words do not appear in the published episode's auto-captions. The episode does discuss the same skill, starting around 00:35:17.

What the full episode adds

A confidence threshold. Between 00:36:02 and 00:37:10 the speaker gives the full rule: "validate the hell out of this, make sure that you have high confidence it actually works. Find some way to test it and verify it and do it. And if you have higher than 8 out of 10 confidence that this will not break anything, just merge it." The clip's version drops both the number and the condition "will not break anything."

A Full Stop skill. The episode describes a sibling: "And then I have full stop, which is everything up until merge. Do not bother me until you can high confidence say this is ready to merge or you can higher confidence say this will never work. Here's why." This skill runs the same pipeline without the merge, which suits teams that require human review.

A second-model review. Before filing the PR, the agent sends the diff to another model: "give these changes to [Sol] with [Codex] in a delegate in T3 code or by calling the CLI... I have it do multiple back and forths until eventually both opus and [Sol] agree or opus has high confidence in the things it's rejecting... Then it files the PR." T3 Code is Theo's open-source agent front end, per a Better Stack guide.

Can an AI agent merge its own pull request on GitHub?

That depends on the repository's settings. GitHub's required-reviews page says "Pull request authors cannot approve their own pull requests." An agent running on your credentials is you, so if the branch requires an approving review, the agent cannot supply it. On those branches Full Send behaves like Full Stop.

The protected-branches page adds a stricter option: the latest push can require approval "by someone other than the person who pushed it."

On a branch that only requires checks, the clip's steps map to two gh commands:

  • "Babysit the pull request until everything passes" maps to gh pr checks --watch, with --fail-fast to stop on the first failure.
  • "Merge it" maps to gh pr merge --auto, which merges "only after necessary requirements are met."

GitHub's auto-merge page describes the same gate: it merges "after all required reviews and status checks pass."

The risk sits in the token. gh pr merge --admin will "merge a pull request that does not meet requirements," and GitHub says repository owners and administrators can merge without an approving review. An agent running with admin rights can take that path, and nothing in the clip's wording prevents it.

Do AI code reviewers count as approvals?

They do not by default. The clip's gate is "all the auto reviewers are happy with it." GitHub's Copilot code review page says Copilot reviews PRs when assigned, and admins can configure automatic reviews. It also says "By default, Copilot's reviews do not count toward required approvals for the pull request." An admin has to enable the setting before a Copilot approval satisfies a required-approval rule. On a default setup, waiting for the bot is a choice made in the prompt.

What is an agent skill?

Anthropic's Agent Skills post, dated October 16, 2025, defines skills as "Organized folders of instructions, scripts, and resources that agents can discover and load dynamically to perform better at specific tasks." In Claude Code each one is a SKILL.md file under .claude/skills/<name>/. agentskills.io lists Claude Code, ChatGPT & Codex, Cursor, Gemini CLI, GitHub Copilot and OpenCode as supporting the format.

Is there a public Full Send SKILL.md?

I found none from Theo. Similar skills exist:

Bottom line

Full Send in the full episode merges only above 8 out of 10 confidence and after a second model reviews the diff. Whether the agent can merge at all depends on your branch protection and the permissions of the token it uses. Teams that require human review should use the Full Stop version.

Sources

Published October 14, 2026. Based on a TikTok by @nerdsnipepod. Full video notes (markdown)