PR

PR

Planner Entry

Create the PR directly in the primary conversation after commit, task-defined checks, and push. Ready PR monitoring and remediation continue through /pr-fixup in the same conversation.

Host detection: This skill works on GitHub, GitLab, and Azure Repos. Detect the host before publication by inspecting git remote get-url origin:

• URL contains dev.azure.com, visualstudio.com, or ssh.dev.azure.com → use the Azure Repos flow below.
• URL contains github.com (or any host you have configured for GitHub) → use the GitHub flow below.
• URL contains gitlab (e.g. gitlab.com, gitlab.acme.corp) → use the GitLab flow at the bottom of this file.
• For self-managed hosts, the user's repository configuration determines the host.

GitHub tool selection: The GitHub flow uses gh CLI by default. If gh is unavailable or fails, including a 401 authentication error, use structured GitHub connector/API tools for PR, check, and review data; authentication failure is unknown state, never clean state. GitLab tool selection: The GitLab flow prefers glab CLI when available; otherwise it shells curl against the REST v4 API using $GITLAB_TOKEN (which the agent runtime injects from the user's secrets store). Azure Repos tool selection: The Azure flow prefers az repos pr create with the Azure DevOps extension. Auth can come from an existing az login session or AZURE_DEVOPS_EXT_PAT.

Available skills

• /commit — Creates the artifact after task-defined checks pass.
• /pr-fixup — Wait for CI checks and CodeRabbit, Greptile, Claude, OpenCode, and cubic review feedback, fix any failures or valid comments, and push.

Options

• --draft — create the PR as draft and skip the fixup step. Use when the work is not ready for review.
• Default (no flag) — create as ready-for-review and continue with /pr-fixup in the same conversation.

Publishing-state precedence: explicit user state (--draft or an explicit ready-for-review request) wins. If the github:yeet plugin is explicitly selected, follow its draft default only when the user did not request either state; otherwise /pr defaults to ready-for-review. Always report the result.

Steps

Track these steps with an internal todo/checklist and mark them complete as you go. Do not create, update, or delete Kandev subtasks for this workflow unless the user explicitly requests task tracking.

1. Uncommitted changes: If there are dirty or staged changes, stop: commit and the affected task checks are required first.

2. Branch: If on main or master, stop and ask the user for a feature branch. Otherwise use the current feature branch as-is.

3. Remote state: Confirm the branch has an upstream and the remote contains the local HEAD. If not, run /push before creating the PR.

   CI artifact bootstrap: If the diff introduces a CI-consumed registry tag or artifact that a workflow on this branch must publish first, follow that publisher's documented bootstrap path before rerunning consumer checks. Request explicit user approval before a workflow_dispatch or other action writes a shared registry, and verify the target tag/digest exists.

4. Screenshots — capture and validate before publication. For a UI-visible change, capture fresh screenshots for every affected viewport before creating the PR. For repository UI, use /playwright-cli with an isolated headless session or the managed apps/web E2E runner. See step 7 for capture routing and blocker classification. When the changed surface is structurally absent on another viewport, record that rationale instead of capturing an unrelated screen. Use synthetic or redacted data, validate the assets, and compress PNGs using the recipe in step 7. If capture is impossible, report the concrete blocker and stop before PR publication. For non-UI changes, record that screenshots are not required and continue.

5. Create the PR. Before creating, check open PRs for the current branch and inspect task-linked PR metadata. Reuse an existing PR; create a duplicate only when the user explicitly requests separate PRs. Use --draft if requested, otherwise create as ready-for-review.

   Architecture and scope gate: Before running any PR creation command, verify the authenticated actor's repository permission. Do not treat a user statement as proof of maintainer status. On GitHub, query gh api repos/{owner}/{repo}/collaborators/{login}/permission; push, maintain, and admin permissions are write-authorized, so those actors may open large or architectural PRs directly. On other hosts, use the equivalent repository permission check. When the host confirms write access, the permission itself satisfies this gate and no linked issue is required solely for this purpose. If permission cannot be verified, or the actor has no write access, require a linked issue with maintainer discussion before opening a large or architectural PR. If that issue or discussion is missing, stop and report the blocker. Do not create an issue or open a PR solely to start the discussion. Prefer one logical change and the smallest practical diff; split unrelated cleanup, refactoring, and feature work into separate PRs.

   PR title must follow Conventional Commits format (see /commit for full rules). CI validates via pr-title.yml — the PR title becomes the squash-merge commit used for release notes.

   PR body must be built from .github/pull_request_template.md; fail fast if it is missing. Read the whole template before writing the body. Treat HTML comments as authoring instructions for the agent, not as output:

   • Fill the template's required sections from the actual diff, commits, and verification performed.
   • Remove optional sections that add no value for this change.
   • Preserve static required sections such as checklists exactly as the template provides them; do not pre-fill unchecked boxes.
   • For docs-only PRs, keep code-centric checklist items unchanged when they do not apply, and list the docs-safe validation commands actually run.
   • Include related issue closing text only when an actual issue number is known.
   • Remove all HTML comments/placeholders from the final body.
   • Do NOT add tool attribution footers.
   • Before creating the PR, self-check that the final body has no <!--, no empty required sections, and no placeholder text.
   • If the diff touches user-visible UI, append the already captured assets in a ## Screenshots section after the PR is created (step 7) — don't add a placeholder for it here.

   test -f .github/pull_request_template.md
   # Build /tmp/pr-body.md from the template, using comments as instructions
   # and removing them from the final file.
   gh pr create [--draft] --title "type: description" --body-file /tmp/pr-body.md

   Do not fall back to hand-composed --body prose. If creation fails, surface the exact stderr, fix the template/body-file problem, and retry with --body-file.

   If gh pr create fails after the branch is pushed with a credential-lease or repository-scope error, use the REST fallback with the same template body. Build the payload with jq --rawfile and submit it with gh api --method POST repos/<owner>/<repo>/pulls --input <payload-file>; preserve the validated title, head, base, and body, and never hand-escape Markdown or JSON.

   After creating the PR through gh pr create or the REST fallback, fetch it with gh pr view <number> --json url,state,headRefName,baseRefName,headRefOid. Verify it is open and its head/base match the intended branch and target before reporting the URL.

6. If ready (not draft): For GitHub, do not begin /pr-fixup until any required screenshot embedding in step 7 is complete.

7. Screenshots — publish already captured assets. If the diff touches user-visible UI (typically under apps/web/, excluding e2e-only or backend-only edits), publish the affected-viewport assets captured and validated in step 4 through the host-specific flow before treating the PR as complete — do not wait to be asked. Preserve any structural-absence rationale recorded in step 4.

   Capture prerequisite:

   • Start repository UI verification with /playwright-cli or the managed E2E runner, rather than a connected personal browser or in-app browser tool. The CLI's default profile is isolated; do not use --extension or a user's persistent browser profile for disposable repository captures.
   • If pnpm --dir apps exec playwright-cli list has no local browser, use the managed apps/web E2E runner with a disposable capture spec instead of treating capture as blocked. Name mobile specs mobile-*.spec.ts, write assets to ignored apps/web/.pr-assets, inspect/compress them, then remove the temporary spec and confirm git status is clean.
   • Classify failures against the capability that failed. A connected-browser saved-site denial does not prove that the repository Playwright workflow cannot capture. Honor that tool's denial and any prohibition on alternate routes; do not evade it by changing hosts, ports, profiles, or providers. If alternate execution needs authorization, explain the boundary once and request it. Filesystem sandbox changes can justify retrying a failed local Playwright launch, but do not clear saved browser permissions. Report a capture blocker only with evidence from the authorized capture workflow.
   • After opening a popover or dialog, assert that the intended surface is visible and await finite active CSS animations (element.getAnimations().finished) before capture; do not publish a mid-transition asset.
   • For disposable capture specs, prefer the existing prCapture fixture from apps/web/e2e/fixtures/test-base.ts; it writes the expected filenames and manifest for PR assets.
   • apps/web/e2e/scripts/run-e2e.sh clears .pr-assets at the start of each managed-runner invocation. Capture desktop and mobile assets in one invocation when possible; if separate runs are necessary, preserve/merge the prior assets and revalidate the complete manifest afterward.
   • Reuse only fresh entries from apps/web/.pr-assets/manifest.json. After every capture, require a non-empty manifest with the intended fresh asset entries: test -s apps/web/.pr-assets/manifest.json. If it is absent or lacks the capture, do not treat the run as successful; rerun with --host and report the managed-runner gap.
   • Start each publication attempt with a fresh PR-scoped capture directory. Before publishing, compare the manifest entries with the explicit expected viewport/capture list and reject stale files or unexpected entries.
   • Before inspecting, compressing, or publishing the assets, verify that every manifest entry maps to an existing file under apps/web/.pr-assets. If any entry is missing, treat the capture as incomplete and restore or recapture the asset; revalidate the complete mapping immediately before publication.
   • If required assets are missing, run the Playwright capture before publication; do not create the PR first.
   • For a mobile capture through pnpm e2e:run, select the runner project before the Playwright separator and use a mobile-*.spec.ts filename: pnpm e2e:run --project mobile-chrome e2e/tests/<area>/mobile-<capture>.spec.ts. --project after -- is only a Playwright argument and leaves the runner on its default Chromium project.
   • In managed Docker E2E, write capture files directly to the mounted workspace path /work/apps/web/.pr-assets/<name>.png, not testInfo.outputPath(...), which is container-local. After the run, confirm the host has the files before inspecting or compressing them.
   • Reject any asset that exposes secrets, authentication tokens, or personally identifiable information and stop for recapture.
   • Compress PNGs before embedding. Prefer a system pngquant; when it is unavailable, use the supported ephemeral fallback rather than skipping compression:

     if command -v pngquant >/dev/null 2>&1; then
       pngquant --quality 65-90 --ext .png --force apps/web/.pr-assets/*.png
     else
       (
         cd apps
         pnpm dlx pngquant-bin@9.0.0 --quality 65-90 --ext .png --force web/.pr-assets/*.png
       )
     fi

   Embed (GitHub only — image binaries must never merge into main): GitHub has no public API to upload images into a PR body (drag-and-drop is web-UI only), so publish the images on an orphan commit that can never be merged and reference them with SHA-pinned raw URLs:

   blob=$(git hash-object -w apps/web/.pr-assets/shot.png)
   printf '100644 blob %s\tshot.png\n' "$blob" > /tmp/tree
   # repeat the hash-object + printf lines, appending one line per file to /tmp/tree
   tree=$(git mktree < /tmp/tree)
   commit=$(git commit-tree "$tree" -m "media: screenshots for PR #<N>")
   git push origin "$commit:refs/heads/media/pr-<N>-screenshots"

   (A quoting glitch can make the first push report failure — retry with the literal commit SHA.) Reference each image in the PR body under a ## Screenshots section using GitHub-Flavored Markdown image syntax and dash-named files (no spaces): ![Short descriptive caption](https://raw.githubusercontent.com/<owner>/<repo>/<media-commit-sha>/shot.png) Bare image URLs render as clickable links instead of previews, so never add a screenshot URL without the surrounding ![alt text](...) wrapper.

   Append the section to the PR body:

   gh pr edit <PR_NUMBER> --body-file <file>

   gh pr edit can fail on this repo (GraphQL touches the deprecated Projects-classic API), or exit successfully after only printing a Projects-classic deprecation warning without changing the body. Treat the read-back as authoritative; if the intended section is absent, fall back to REST — build the payload with jq --rawfile, never by hand-escaping shell strings:

   set -euo pipefail
   PAYLOAD="/tmp/pr-body-<PR_NUMBER>-payload.json"
   jq -n --rawfile body "<body-file>" '{body: $body}' > "$PAYLOAD"
   jq empty "$PAYLOAD"
   gh api --method PATCH repos/:owner/:repo/pulls/<PR_NUMBER> --input "$PAYLOAD" --silent

   For a write-only PATCH, --silent prevents gh from decoding a response body that the caller does not consume. Redirecting stdout to /dev/null does not prevent a truncated or empty JSON response from making gh report an error after the mutation succeeds. Read the resource back with a separate GET and verify the intended change. JSON payloads: Use jq (or an equivalent JSON tool) to build and validate payloads without interpolating untrusted Markdown into shell syntax. When command output is redirected, piped, or consumed by jq or another parser, run the command through rtk proxy so stdout is byte-preserving; normal RTK filtering can truncate or annotate machine-readable output. Keep the REST fallback byte-preserving for command substitutions, xargs, and any other consumer that expects unmodified Git or JSON output.

   Preserve the existing PR description: The PR body is a shared, mutable document. Preview automation and other bots may add sections after the PR is created, so never PATCH a body reconstructed from the creation-time template or a stale /tmp/pr-body.md. Before every post-creation update:

   This live-body procedure applies to every post-creation body edit, not only screenshots. For each generated block, use operation-owned markers and verify that its start and end markers occur exactly once before and after the update.

   1. Fetch the current body from GitHub and keep that pristine live response as the merge base: gh pr view <PR_NUMBER> --json body --jq .body > /tmp/pr-body-latest.md.
   2. Change only the section owned by this operation. For screenshots, wrap the section in <!-- kandev-screenshots-start --> and <!-- kandev-screenshots-end --> markers (or replace the existing ## Screenshots block if those markers are not present). Preserve every byte outside that range, including <!-- kandev-preview-start --> ... <!-- kandev-preview-end -->.
   3. Build a separate merged candidate, then re-fetch the live body immediately before PATCH and compare the two live snapshots (not the candidate with a snapshot). Use PR-scoped temporary filenames, fail immediately on a differing cmp, and discard any payload on failure; never reuse a prior /tmp payload. Extract JSON bodies byte-for-byte with jq -j .body before cmp; jq -r .body appends a newline and can report a false mismatch. Verify the payload contains this PR's current body and required sentinels before PATCH. If it changed, re-fetch and merge again; do not overwrite the newer body.
   4. After PATCH, read the body back and verify both the intended change and all previously present sentinel sections are still present. For edits to evidence or validation text, compare claims about files, tests, and commands with the latest verification and correct stale counts before reporting the PR. A body mutation can start fresh documentation/check work and invalidates the prior PR snapshot, so refresh pr-state and pr-resolve and rerun the appropriate pr-await wait before treating the PR as complete.

   The REST PATCH endpoint replaces the complete body and does not provide a convenient description-level compare-and-swap, so this fetch/merge/check sequence is required even when the edit appears small.

   Before treating screenshot publication as complete, inspect the submitted body and verify every screenshot entry is a Markdown image embed (![...](https://raw.githubusercontent.com/.../*.png)) rather than a bare URL. Use gh pr view <PR_NUMBER> --json body --jq .body for this check.

   Never commit the screenshot binaries to the PR branch itself — only to the throwaway media/pr-<N>-screenshots ref (git rm them from the PR branch tip if they were committed there earlier; with squash-merge, deleting at tip is enough). The docs/screenshots/ directory is for product/docs imagery that is meant to merge — don't confuse the two. The media branch must survive branch-cleanup sweeps; deleting it 404s the images in the PR body, so don't treat "unmerged branch" as automatically safe to delete.

8. Report the PR URL after all applicable steps are complete. For a ready GitHub PR, continue with /pr-fixup in the same conversation when requested.