yolo
Ship the current branch end-to-end without stopping to ask — commit staged/unstaged work, push, open a PR, watch CI, merge once green, and where merging…
Skill by bonzofenix
Chalk Labs
Need help deploying yolo?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/bonzofenix-yolo
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: devops, data
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/yolo
What's inside
- Skill name
yolo- Description
- Ship the current branch end-to-end without stopping to ask — commit staged/unstaged work, push, open a PR, watch CI, merge once green, and where merging deploys, follow it to production and confirm the new build is actually serving. Use when the user says "yolo", "ship it", "commit push and merge", "just merge it", or otherwise wants the whole commit→PR→merge→deploy pipeline run autonomously. Do NOT use for review-only requests or when the user only wants a commit.
- License
- MIT
- Instructions
- 333 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# yolo — commit, push, PR, watch CI, merge, verify in production
Run the full ship pipeline autonomously. **Done means running in production**,
not merged: where merging triggers a deploy, the run is not finished until the
new build is confirmed serving (step 7). The point of `yolo` is that the user
has opted into NOT being asked at each step — proceed through all stages without
pausing for confirmation, stopping ONLY on a hard failure (merge conflict, red
CI that isn't a known flake, a protected-branch rejection).
## Preconditions (check fast, fail loud)
1. Confirm you are in a git repo and NOT on the default branch (`main`/`master`).
- If on the default branch, create one first:
`git switch -c <short-kebab-feature-name>` derived from the change.
- If a worktree/session branch already exists, use it.
2. `gh auth status` must be usable. If `gh` is missing/unauthed, stop and tell
the user to auth (`gh auth login`) — do not fall back to raw API.
## Pipeline
### 0. Commit FIRST, then review
Commit before review, not after. Review agents run in this same working tree,
and a dirty tree is shared mutable state between you and them.
This is not hypothetical. On a real run a review agent tried to clean up its own
scratch files, hit a permission denial on `rm`, fell back to
`git clean -f internal/store/`, and destroyed five untracked files — two of them
the main thread's in-progress tests, unrecoverable because they were never
committed. A second agent reverted an uncommitted edit that was mid-write.
Committing first makes agent scratch visibly untracked, makes `git clean`
survivable, and means the review reads a stable snapshot rather than a tree
changing underneath it.
- `git add -A` (stage everything) unless the user scoped specific paths.
- Write a real Conventional Commits message (`type(scope): subject`, imperative,
≤50-char subject; body only when the "why" isn't obvious). Never a placeholder.
- If nothing to commit AND the branch is already pushed, skip to step 2.
### 1. Review — mandatory, never skipped
Review is the reason to run yolo, not a preamble to it. This is the only review
the change is guaranteed to get — no required status check gates merges in this
repo, and other repos may have none either. Never assume GitHub will catch what
you skip.
- Run `/pr-review-toolkit:review-pr` (all aspects) against the committed diff
(`git diff origin/<default-branch>...HEAD`).
- **Tell the agents they may read and report only.** They must not delete, move,
or modify any file, and must not run `git clean`, `git checkout --`,
`git stash`, or `git restore`. Scratch files they create are to be left in
place and named in their report so the main thread can remove them. An agent
that "tidies up" is editing a tree someone else is working in.
- Run it for **every** diff. Do not skip because the change looks small, is
docs-only, is prose, or "obviously" has no executable code — that judgment is
not yours to make here, and the cost of a needless review is far lower than the
cost of an unreviewed change reaching `main` unchecked.
- **State the stakes in the prompt.** Findings track what the agents are told to
look for: "review this diff" surfaces style, while naming what the change can
destroy ("this migration runs against a live database holding 318
irreplaceable rows") surfaces data loss. Say what is irreversible, what is
production, and what cannot be regenerated.
- Apply what it recommends: fix Critical + Important issues directly. Suggestions
are optional — apply if cheap/obvious, skip otherwise. Amend or add commits as
you go; the branch is already committed, so fixes are ordinary commits.
- Don't stop to ask before applying fixes — that's the point of yolo. Only stop
if a finding is ambiguous enough that guessing the fix risks breaking behavior.
- Re-run is not required after fixes; proceed straight to push.
- If the review cannot run, or does not return findings you can act on (including
a `max_turns` timeout or a half-completed pass), treat it as not having run:
say so plainly and STOP. Never merge something that was never reviewed while
implying it was.
### 1b. Schema migrations — run it against real data before shipping
If the diff touches `migrations/` (or otherwise changes a schema), reading it is
not enough. Both of the worst bugs found on a real run were invisible in the
diff and only appeared when the migration was executed:
- **A cascade nobody mentioned.** Foreign keys were on, so `DROP TABLE videos`
during a table rebuild fired `ON DELETE CASCADE` and silently emptied
`video_index_notes` — a table the diff never names, joined by every search.
No error, and the scanner that could have rebuilt it was deleted in the same
change.
- **A plausible value that aborts the deploy.** One teacher name appearing twice
made a name-join fan one row into two, failing on
`UNIQUE constraint failed: videos_new.id` — naming neither the table at fault
nor the duplicate.
So:
- Copy the production database and run the migration against the copy. **Copy
the `-wal` and `-shm` files too** for SQLite — a copy of the `.db` alone reads
stale committed state and will show you a migration that "did not run".
- Assert **row conservation on every table**, not just the one being altered.
Count before, count after, compare. That single check catches both bugs above:
a cascade is a count that dropped, a fan-out is a count that grew.
- Check the invariants the new schema claims. If a column caches something
derived, verify no row disagrees with the function that derives it.
- Report the before/after table in the PR body. It is the evidence the migration
is safe, and the reviewer cannot reproduce it.
Add a regression test that seeds the tables the migration does *not* mention.
A test that only seeds what the diff touches will pass while a neighbouring
table is being emptied.
### 2. Push
- `git push -u origin HEAD`.
### 3. PR
- If a PR for this branch already exists (`gh pr view --json number,url,state`),
reuse it. Otherwise `gh pr create --fill` (or a written title/body summarizing
what and why). Base = the repo default branch unless told otherwise.
- Capture the PR number + URL; report the URL to the user.
### 4. Watch CI
- First check whether there is anything to watch: `gh pr checks <num>`. Checks can
take a few seconds to register after a push, so if it reports none, wait ~15s andFrom bonzofenix/skills, MIT.
More from bonzofenix/skills
- 0.566score
whatsapp-to-db
bonzofenix
Turn a WhatsApp chat export (.zip, unzipped folder, or bare chat .txt) into a queryable SQLite database — every message, every voice note transcribed locally…
SkillAFreshness: AClaude Code +9 - 0.560score
claude-new
bonzofenix
Start a new background Claude Code session in a project under ~/workspace, so it shows up in the agent view.
SkillAFreshness: AClaude Code +9 - 0.557score
analyze-ci
bonzofenix
Analyze CI checks for current branch's PR - polls pending checks until complete, diagnoses failures, auto-fixes straightforward issues, and suggests fixes for…
SkillAFreshness: AClaude Code +9 - 0.554score
unresolved-comments
bonzofenix
Fetch and display only unresolved comments from a GitHub pull request
SkillAFreshness: AClaude Code +9 - 0.554score
smart-pr
bonzofenix
Generate or update PR title and description based on branch diffs
SkillAFreshness: AClaude Code +9 - 0.554score
last-screenshot
bonzofenix
References the most recent screenshot from the user's Desktop for analysis
SkillAFreshness: AClaude Code +9 - 0.554score
sonarcloud-issues
bonzofenix
Fetch and display SonarCloud/SonarQube analysis issues for a GitHub PR
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Sep 25, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/bonzofenix/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from bonzofenix
- 0.566score
whatsapp-to-db
bonzofenix
Turn a WhatsApp chat export (.zip, unzipped folder, or bare chat .txt) into a queryable SQLite database — every message, every voice note transcribed locally…
SkillAFreshness: AClaude Code +9 - 0.560score
claude-new
bonzofenix
Start a new background Claude Code session in a project under ~/workspace, so it shows up in the agent view.
SkillAFreshness: AClaude Code +9 - 0.557score
analyze-ci
bonzofenix
Analyze CI checks for current branch's PR - polls pending checks until complete, diagnoses failures, auto-fixes straightforward issues, and suggests fixes for…
SkillAFreshness: AClaude Code +9 - 0.554score
unresolved-comments
bonzofenix
Fetch and display only unresolved comments from a GitHub pull request
SkillAFreshness: AClaude Code +9 - 0.554score
smart-pr
bonzofenix
Generate or update PR title and description based on branch diffs
SkillAFreshness: AClaude Code +9 - 0.554score
last-screenshot
bonzofenix
References the most recent screenshot from the user's Desktop for analysis
SkillAFreshness: AClaude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get yolo running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
yolo skill. Agent Mart. https://agentmart.dev/skills/bonzofenix-yolo (last updated 2026-10-24)