adhd-output-style
This skill should be used when the user asks for "ADHD output", "fewer output tokens", "short numbered steps", "limited working…
This skill should be used when user asks to "create a PR", "make a pull request", "open PR for this branch", "submit changes as PR", "push and create PR", or explicitly invokes "create-pr".
Claude Code
$ npx -y skills add fcakyon/claude-codex-settings --skill create-pr --agent claude-code
Codex
$ npx skills add fcakyon/claude-codex-settings --skill create-pr -a codex -y
Cursor
$ npx skills add fcakyon/claude-codex-settings --skill create-pr -a cursor -y
Gemini CLI
$ npx skills add fcakyon/claude-codex-settings --skill create-pr -a gemini-cli -y
GitHub Copilot
$ npx skills add fcakyon/claude-codex-settings --skill create-pr -a github-copilot -y
OpenCode
$ npx skills add fcakyon/claude-codex-settings --skill create-pr -a opencode -y
Windsurf
$ npx skills add fcakyon/claude-codex-settings --skill create-pr -a windsurf -y
name: create-pr description: This skill should be used when user asks to "create a PR", "make a pull request", "open PR for this branch", "submit changes as PR", "push and create PR", or explicitly invokes "create-pr".
Create a PR a reviewer can understand without reading the conversation. Use extra invocation text and session findings to explain the problem, final behavior, and evidence. Follow the user's requested format and the target repo's guidance.
These rules apply to pull requests. For problem-focused issue titles and structured issue reports, use [create-issue](../create-issue/SKILL.md).
1. **Inspect the scope and base**
branch already contains the requested changes.
fetch that remote and fast-forward the local base with `git merge --ff-only <remote>/<base>`. Preserve dirty work and stop if fast-forwarding fails. Do not reset it away.
2. **Finish and inspect the change**
commits included. Follow `commit-staged` for any staged changes.
actual base and head, or `gh pr diff <pr>` if this checkout is on another branch.
the final change, including links, counts, errors, and benchmark provenance.
3. **Write and publish**
rather than opening a duplicate.
Never interpolate text into a shell command or include command output by accident.
`gh pr list --repo <owner>/<repo> --author @me --limit 5 --json reviewRequests`.
Check code fences, tables, links, and preserved media. For visual changes, open the PR and verify the images render. Report the URL and any incomplete verification.
`Fix copied session text losing line breaks`. No internal brand names or vague claims. Add a Conventional Commits prefix (`fix: ...`) only when the repo squash-merges, since the title becomes the single commit, and match its recent merged titles. Otherwise use no type prefix. Rewrite the title and body when the final scope changes.
does the fix matter? For a feature, show what the user can now do. Skip "This PR" openers.
paragraph or cut useful evidence to satisfy a word count. Small changes may need only one sentence and a snippet. Use short sections when reproduction, fix, and results need separation. Omit empty or boilerplate sections.
behavior, then explain how the final change fixes it. Use exact logs where they show the failure. Skip setup noise, raw tool output, and the debugging diary.
or separately labeled fenced Before and After outputs. Prefer these to red/green diff lines that force the reviewer to infer the outcome. Do not force snippets onto copy edits.
workload, or sample count needed to interpret them. Link relevant issues, commits, docs, and result artifacts where they support the claim, rather than listing changed files.
material limitations or pre-existing failures. A future test checklist is not evidence. Do not claim full-run success from a focused check, or understate verified full runs.
counts or base-branch noise when they help distinguish the actual review scope.
viewport and scenario, and inspect the visuals before using them. Analysis/plot changes need representative outputs and measured results readable at the displayed size.
repo's hosting rules. Where release uploads are permitted, use an existing appropriate release and embed its asset URLs. Preserve existing GitHub attachment URLs as well.
paths, the PR link, and the sentence/table cell where each image belongs for manual upload. Do not invent attachment URLs or claim pending uploads are visible.
Battle-tested Claude Code, OpenAI Codex, Cursor configs, plugins, hooks and agents with Kimi, MiniMax and GLM API support.
This skill should be used when the user asks for "ADHD output", "fewer output tokens", "short numbered steps", "limited working…
This skill should be used when the user asks to "build an AEO strategy", "improve AI search visibility", "track buyer prompts",…
Agent-browser usage guide. Read this before running any agent-browser commands. Covers the snapshot-and-ref workflow, navigating…
Reverse-engineer a website's internal API by recording browser traffic into a HAR file, then generate a standalone client or CLI…
Systematically explore and test a web application to find bugs, UX issues, and other problems. Use when asked to "dogfood", "QA",…
You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or…
Ask which skill or flow fits your situation. A router over the skills in this repo.