[mikepitre]

What & why

• CodeRabbit reviews a pull request when it opens or turns ready, and its device evidence check judges the description it reads then. The skill had the agent open the PR ready and run attach right after, so the review could start before the evidence was written. On [BLACKSMITH TEST] fix(auth): shorten the phone field switcher label to "Use email or username" clerk-ios#805 the table landed 16 seconds after the PR opened, CodeRabbit had already read the earlier description, and the check failed until someone asked for a fresh review.
• The skill now says to open the PR with gh pr create --draft, run attach and down, and run gh pr ready <n> once the description holds the evidence. For a remote session, where the evidence arrives about a minute after down, it says to check that gh pr view <n> --json body shows the verify-evidence block first. references/remote.md says the same.
• team-review-request and code-owner-approval already run on ready_for_review, and local attach and the verify-attach workflow both publish to drafts, so nothing else changes.
• In a Claude Code cloud session GitHub's GraphQL API is blocked, so gh pr create, gh pr view and gh pr ready fail there. The skill names the REST forms that work, including POST repos/<repo>/pulls/<n>/ccr/ready_for_review, which a cloud session on clerk/clerk-ios confirmed by flipping test PR Update dependency com.materialkolor:material-kolor to v5 #805 to draft and back. The wait for the evidence polls the description over REST for the table row and gives up after ten minutes.

Testing

• Docs only. The next PR an agent opens with this skill shows whether CodeRabbit's first review passes the device evidence check.

Risk

• Low: docs only. A PR stays a draft if an agent stops before gh pr ready, which is visible and easy to fix by hand.

How to revert

• A plain revert is enough.

Screenshots / video

• None. Docs only.

🤖 Generated with Claude Code

[coderabbitai]

No actionable comments were generated in the recent review.

Before this change, the verification workflow attached device evidence but did not require the pull request to remain a draft until the evidence appeared. It now opens the pull request as a draft and marks it ready only after the evidence table is present in the description. For remote sessions, the workflow checks for the table for up to ten minutes after down; Claude Code cloud sessions use documented REST alternatives when GraphQL-backed gh pr commands fail. Reviewers should start with the workflow and evidence check in SKILL.md, then read references/remote.md for remote-session publishing details.

Walkthrough

The verification instructions create pull requests as drafts, check for the evidence table, and mark them ready after it appears. They add polling after remote teardown and document GitHub API alternatives for Claude Code cloud sessions.

Changes

Verification evidence workflow

Layer / File(s)	Summary
Draft pull request until evidence is present .claude/skills/verify-clerk-android/SKILL.md, .claude/skills/verify-clerk-android/references/remote.md	The workflow creates a draft pull request before attaching evidence, waits for the evidence table to appear, and then marks the pull request ready. It documents GitHub API alternatives for pull-request commands in Claude Code cloud sessions.

Estimated code review effort: 1 (Trivial) | ~5 minutes

Change: Other

Possibly related PRs

• clerk/clerk-android#1190: Changes how the evidence publisher updates the pull request description that this workflow checks.
• clerk/clerk-android#1192: Adds a code-owner approval workflow that skips draft pull requests.
• clerk/clerk-android#1198: Adds a workflow triggered when a pull request becomes ready for review.

Suggested reviewers: wobsoriano

Merge Risk: Moderate · up to e69da

The verification workflow can mark a pull request ready before device evidence is available, so this existing review risk should be resolved before merge.

Pre-merge checks: 6

Check name	Status	Explanation
Title check		The title clearly identifies the documentation change and its purpose: keeping the pull request in draft status until evidence is available.
Description check		The description includes all required sections and provides clear context, testing status, risk, and rollback information. The change is documented as docs-only, so omitting What to focus on is accept…
Linked Issues check		Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check		Check skipped because no linked issues were found for this pull request.
Testing, Risk, And Revert		The PR description includes all three required sections. Testing states that the change is docs-only and identifies the next PR as the validation point. Risk states that the risk is a PR remaining dra…
Device Evidence		The pull request changes only .claude/skills/verify-clerk-android/SKILL.md and references/remote.md. It does not change any file under source/api/, source/ui/, or e2e/, so device evidence is…

Full details: Description check

Explanation

The description includes all required sections and provides clear context, testing status, risk, and rollback information. The change is documented as docs-only, so omitting What to focus on is acceptable.

Full details: Testing, Risk, And Revert

Explanation

The PR description includes all three required sections. Testing states that the change is docs-only and identifies the next PR as the validation point. Risk states that the risk is a PR remaining draft if the agent stops early. How to revert states that a plain revert is sufficient.

Full details: Device Evidence

Explanation

The pull request changes only .claude/skills/verify-clerk-android/SKILL.md and references/remote.md. It does not change any file under source/api/, source/ui/, or e2e/, so device evidence is not required by this check.

CodeRabbit Verify

• Run CodeRabbit Verify to test the changes in this pull request.

Recent review info

Automatic approval policy

Commit: e69daccc598d.

Setting	Effective value	Source
Automatic approvals	Enabled	Global configuration
Merge risk	Moderate	Global configuration
Security risk	Not set	Global configuration

Not set means no limit was configured after inheritance. Use @coderabbitai configuration to inspect the resolved settings and their sources.

Run configuration

• Source mode: Auto
• Repository: clerk/clerk-android
• Branch: verify-draft-until-evidence
• File: .coderabbit.yaml
• Commit: e69dacc
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Team
• Run ID: 994c64d8-f234-427c-95ec-14c7ddf0652a

Commits

Reviewing files that changed from the base of the PR and between c16c82d and e69dacc.

Files selected for processing: 1

• .claude/skills/verify-clerk-android/SKILL.md

Limit details: You’ve used all 5 included reviews currently available. Your 57 included PR review attempts over the past 7 days set your current allowance at 5 reviews per hour.

CodeRabbit: Dismissed by CodeRabbit