Index of every page on this site: llms.txt
All skills
ccusage avatar

/create-pr

@7e789e8
by ccusageccusage/ccusage20k stars
934
Run it yourself

Runs the full PR lifecycle. Use when creating a branch, committing, pushing, opening a PR, requesting AI review, replying to review threads, and driving CI to green.

Use this Skill: https://skilld.dev/gh/ccusage/ccusage/create-pr

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈44 tokens for metadata: the name and description. ≈670 when used: this file. ≈800 more on demand in 3 files.

Description uses 2.2% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈177 characters, or 2.2%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Create PR

Split independent work into separate PRs and stack dependent work.

Workflow

  1. Branch. Pushing to main needs explicit permission from the user; everything else goes through a feature branch named after the change (existing branches read <type>/<description>). Commits come from the commit skill, so they stay atomic and independently revertable. Leave unrelated formatter churn, generated files, and user edits outside the task unstaged.
  2. Push and open the PR — references/open-pr.md. Use the Conventional Commit scope rules in the commit skill.
  3. Request and handle AI review — references/ai-review.md, with the gh reply and thread-state calls in references/gh-review.md.
  4. Watch CI after opening and after every push with gh pr checks. Read the failing step logs and annotations rather than the summary, then use the fix-ci skill for the fix, and ask the bots for another pass if they do not rerun on their own.
  5. Merge only when the user explicitly asks and the conditions below hold: gh pr merge <pr> --squash, without --delete-branch or --body-file. When the PR fixes an issue someone else reported, credit that reporter with a Co-authored-by trailer on one of the PR's commits before merging, using their noreply address <id>+<login>@users.noreply.github.com (gh api users/<login> --jq .id): the squash body is built from the commit messages.
  6. After merging, reply on each linked issue and on any original contributor PR it superseded. For a branch offered without a PR, reply where it was offered. Say what landed and in which PR, and, when it changes the published package, the release that shipped it or that it ships with the next one. Do this in particular when someone pinged or offered a branch and got no answer.

Ready means

  • The branch is pushed, the PR exists, and its body describes the change and the validation that ran.
  • CodeRabbit — and Cubic when it is available on the PR — reviewed the latest pushed commit, with no unresolved actionable feedback.
  • Every required check passes. Queued, cancelled, failing, or missing required checks all mean not ready.
  • The user has the PR URL plus any residual risk or pending external state.

When a bot or CI system stays silent past a reasonable polling window, state exactly what is pending instead of claiming completion, and leave the visible PR comment or CI state for follow-up.

Context

  • Reviewer bots act on a comment only when their handle is mentioned — in the initial request and in every reply asking them to do something.
  • Amending or force-pushing after reviewers have read the PR needs an explicit request from the user.

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

1 warning3d3 checks · Risk SAFE
  • Gen Agent Trust Hub4d

    This skill automates the full GitHub pull request lifecycle, including branch management, PR creation, AI-driven code reviews, and merging. It incorporates security best practices such as requiring explicit user permission for merges and pushes to protected branches. The primary security consideration is the potential for indirect prompt injection from processing external PR comments and reviews.

  • Socket3d

    No alerts

  • Snyk4d

    Risk: MEDIUM · 1 issue

Signed by skilld at 7e789e8. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated yesterday

README badge

README badge for ccusage/ccusage/create-pr