pr
Create or reuse a ready-for-review GitHub PR for the current repository change against the requested base, or the repository's remote default branch when none…
Skill by toy-crane
Chalk Labs
Need help deploying pr?
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/toy-crane-pr
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
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 and code. Checked Oct 26, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 2
codeload.github.com· found in install commandgithub.com· found in code
Files it touches
~/.claude/skills/pr
What's inside
- Skill name
pr- Description
- Create or reuse a ready-for-review GitHub PR for the current repository change against the requested base, or the repository's remote default branch when none is named. Always use this skill for an actual PR publication, including requests to create, open, raise, publish, or reuse a pull request, put work up for review, or make and share a GitHub review link. Complete the needed commit, base synchronization, and branch publication, but stop before merge.
- License
- MIT
- Instructions
- 114 lines in SKILL.md · view on GitHub
Files (6)
- SKILL.md
- openai.yaml · agents/
- control-trigger-evals.json · evals/
- evals.json · evals/
- heldout-trigger-evals.json · evals/
- trigger-evals.json · evals/
SKILL.md (first lines)
# Create a pull request Turn the current request's change into one ready-for-review pull request against the named base, or the remote's advertised default branch when none is named. Complete the necessary local Git work instead of requiring other skills to be installed. Preserve unrelated work while committing the request's changes as logical Conventional Commits. Keep credentials and other sensitive files out of the history, honor repository checks, and move work on a detached or protected checkout to a suitably named branch. Base the branch on the fetched target and publish it without overwriting unexpected remote work. Reuse an existing pull request for the same head and base; when the change is already represented on that remote base, report the state instead of duplicating it. Create the pull request ready for review with a title and body describing the actual change. When the repository's `AGENTS.md` or `CLAUDE.md` carries an `## Issue tracker` section, read it and its linked detailed convention when present; legacy inline conventions also work without the setup skill installed. Without the section, skip every issue step below and leave the body unchanged. ## Link each spec folder to one issue With a tracker configured, gather the spec folders this PR carries before creating it: every `docs/specs/<slug>/` folder the branch adds or changes, and every folder named by a `Spec-Folder: docs/specs/<slug>/` commit trailer, that still exists at the head. Skip folders the branch deletes. A folder's `spec.md` names its issue with an `Issue: <tracker>:<id>` line; use that ID and create nothing. For each gathered folder without that line, create one issue through the convention's create operation: the title is the first heading of `spec.md` with no prefix, the body is one spec section rendered from `spec.md` between the convention's spec section markers, and the issue carries the convention's review signal so general triage leaves it alone before the PR links it. The section holds the first section's text, the remaining section headings, the folder path, and the issues of folders its `Blocked by: docs/specs/<other>/` lines name, read from their own `Issue:` lines; report a blocker folder that has none. Then write `Issue: <tracker>:<id>` in the convention's ID form beside the spec's other links, commit it on the same branch, and only then publish and open the PR. Never search the tracker by title to find or reuse an issue. If the issue was created but the line could not be committed, stop before opening the PR and report the created ID so a rerun records it instead of creating another. Put an `Issue: <tracker>:<id>` line in the PR body for each linked issue, so a later pass can find the PR after the branch is gone. A PR with no spec folder that fixes a known issue directly carries the same line. Put the convention's closing reference in the body only when the PR actually delivers that issue's implementation, as established by the diff and verification; a trailer or `Issue:` line alone does not establish delivery. A spec-only PR links its issue without a closing reference. When reusing an existing PR, apply the same steps and bring its body in line. Leave merging, required reviews, and release decisions outside this skill's authority. If a linked convention cannot be read, report the missing information rather than guessing issue operations or claiming that the issue was linked or closed. ## Make the body understandable Write for a reviewer without the conversation history: explain the problem, concrete before-and-after behavior, why it changed, and what verification establishes. Scale the explanation to the change; a typo fix needs only a brief description and relevant validation. Include mechanisms, a diagram for complex flows, actual alternatives and trade-offs, affected users or integrations, and specific unresolved judgments when they help assess the change. Keep the whole change understandable beyond those highlighted judgments; omit empty sections and invented alternatives or questions. For screen changes, let the reviewer watch the behavior they are judging. Record the actual base and head running the same flow at matching viewport, data, and state, or reuse evidence verified to match those revisions, and embed the before-and-after videos with a short explanation of the difference and its reason. When motion adds nothing to the judgment, as with copy, styling, or static layout, compare screenshots side by side instead; when such fine visual detail accompanies a behavior change, add that still comparison to the videos. A screen with no before state gets the head video alone. If video cannot be recorded, made safe, or kept within the attachment limit after trimming to the judged behavior, compare screenshots and say why video is missing. Video takes no alt text, so label each one with its revision, flow, and conditions, and identify unavoidable differences between the captures. Separate observed results from source inference and unverified states; a prototype is not runtime evidence. When updating an existing PR, bring its explanation and evidence into line with the final change, replacing outdated visual claims. Before uploading, inspect screenshots and videos for credentials, personal data, and private information. Use safe seeded data or redact those details throughout the media, keeping comparison conditions and the relevant change visible. Upload only the inspected, safe media as GitHub attachments, using a supported mechanism in the current environment. For GitHub CLI, check attachment support: `gh pr create` and `gh pr edit` accept repeatable `--attach` on supported versions and rewrite matching local media references in `--body-file` to uploaded URLs. A Markdown table can place two images side by side. Reference each video as `` alone in its own blank-line-separated paragraph so it renders as a player; a reference inside other text renders as a link, and an unreferenced attachment lands at the end of the body. Keep review captures outside repository history; no separate media host is needed. If baseline execution, capture, or upload is unavailable, state the exact limit and present only the evidence obtained. Inspect the resulting remote body and attachments before reporting them as available to reviewers; a local path is not a published attachment. A partial upload may still create the PR: inspect and repair that PR instead of duplicating it. Carry successful attachment URLs from the remote body into its replacement, and replace unusable local references with an honest limitation if recovery fails. ## Finish Finish when the remote pull request exists in ready state and the user has its URL, base, head branch, and any checks or evidence that remain outstanding.
From toy-crane/skills, MIT.
More from toy-crane/skills
- 0.609score
clean-branches
toy-crane
Clean up the local leftovers of finished work in one pass, removing worktrees and local branches whose pull request merged or closed, detached worktrees…
SkillAFreshness: A★ 3Claude Code +9 - 0.609score
resolve-follow-ups
toy-crane
Resolve evidence-backed `docs/follow-ups/*.md` through a bounded manual, Codex Scheduled task, or Claude Routine sweep.
SkillAFreshness: A★ 3Claude Code +9 - 0.609score
review-sessions
toy-crane
Review this machine's recent Claude Code and Codex session logs for work agents redo by chance every session, and report candidates to hand to a deterministic…
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
babysit-specs
toy-crane
Revise active `docs/specs/SLUG/` specs so they match what has shipped since they were written.
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
implement
toy-crane
Implement or resume settled work from a selected spec folder as one handoff bundle.
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
maintain-project-context
toy-crane
Clean, compact, and reconcile durable project context and any configured issue-tracker mirror without changing established meaning.
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
project-knowledge
toy-crane
Maintain a project's canonical terms and settled decisions that future work should reuse, and record follow-ups a session discovers but does not resolve.
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
expo-dev-loop
toy-crane
Verify Expo and React Native changes in a running app with agent-device.
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
expo-smoke-test
toy-crane
Verify an Expo or React Native change together with the app's core loop on both iOS and Android using agent-device.
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
build-prototype
toy-crane
Build a complete dummy-data HTML prototype or a temporary comparison of up to three alternatives for a specific user situation.
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
merge
toy-crane
Carry the current repository change through a verified GitHub pull request merge into the requested base, or the repository's remote default branch when none…
SkillAFreshness: A★ 3Claude Code +9 - 0.564score
add-stack-context
toy-crane
Audit and establish current agent context for the technologies that define a project's stack.
SkillBFreshness: B★ 3Claude Code +9
Signals
- Stars
- 3
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 8, 2026 (22 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/toy-crane/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from toy-crane
- 0.609score
clean-branches
toy-crane
Clean up the local leftovers of finished work in one pass, removing worktrees and local branches whose pull request merged or closed, detached worktrees…
SkillAFreshness: A★ 3Claude Code +9 - 0.609score
review-sessions
toy-crane
Review this machine's recent Claude Code and Codex session logs for work agents redo by chance every session, and report candidates to hand to a deterministic…
SkillAFreshness: A★ 3Claude Code +9 - 0.609score
resolve-follow-ups
toy-crane
Resolve evidence-backed `docs/follow-ups/*.md` through a bounded manual, Codex Scheduled task, or Claude Routine sweep.
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
build-prototype
toy-crane
Build a complete dummy-data HTML prototype or a temporary comparison of up to three alternatives for a specific user situation.
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
implement
toy-crane
Implement or resume settled work from a selected spec folder as one handoff bundle.
SkillAFreshness: A★ 3Claude Code +9 - 0.604score
expo-smoke-test
toy-crane
Verify an Expo or React Native change together with the app's core loop on both iOS and Android using agent-device.
SkillAFreshness: A★ 3Claude 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 pr 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
pr skill. Agent Mart. https://agentmart.dev/skills/toy-crane-pr (last updated 2026-10-24)