Repository navigation
[#324]: Add /start-issue delivery loop and shared agent command surface - #325
Merged
mattrace-buildmidwestern merged 3 commits intoAug 13, 2026
Merged
Conversation
Give agents a single issue-to-PR path that matches Fluent Mobile board, branch, and Refs conventions.
Contributor
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (12)
📝 WalkthroughWalkthroughThe change establishes Claude as the canonical agent workflow source, adds Cursor command shims, introduces agent skills and architecture hooks, and updates repository guidance for issue delivery, pull requests, onboarding, testing, and synchronization. ChangesAgent tooling and delivery workflows
Estimated code review effort: 4 (Complex) | ~45 minutes Possibly related issues
Possibly related PRs
Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Warning There were issues while running some tools. Please review the errors and either fix the tool's configuration or disable the tool if it's a critical failure. 🔧 ESLint
ESLint install failed: dependency version conflict. Check your lock file or package.json. Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
6 tasks
mattrace-gloo
marked this pull request as ready for review
August 13, 2026 03:34
mattrace-gloo
requested review from
B3RN153 and
JonathanSeehagen
as code owners
August 13, 2026 03:34
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 17
🧹 Nitpick comments (1)
.claude/skills/write-test/SKILL.md (1)
33-37: 🩺 Stability & Availability | 🔵 Trivial | ⚡ Quick winDocument the
fetchmock for service tests.The service row covers
FluentAPI, sync, and storage tests, but the mocking rule only namesjest.mock(...).fetchis a global, so this text does not define how to prevent network I/O in endpoint tests. Add a service rule to mockfetchand restore it after each test, following neighboring service tests.Based on learnings, unit-test new endpoints with mocked
fetchand follow neighboring*.test.tsfiles.Also applies to: 53-55
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.claude/skills/write-test/SKILL.md around lines 33 - 37, Update the service-testing rules in the skill document to explicitly require mocking global fetch for endpoint unit tests and restoring the mock after each test, following neighboring service *.test.ts patterns. Keep the existing per-file jest.mock guidance and live API opt-in instructions unchanged.Source: Learnings
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.claude/agents/code-reviewer.md:
- Around line 99-101: Update the fenced example in the documentation format near
the diff to specify the text language identifier, changing the opening fence to
```text while leaving the example content unchanged.
In @.claude/commands/create-pr.md:
- Around line 114-122: Require a valid issue number from the branch before
continuing, and remove the branch-slug title fallback so PR creation stops when
none is found. Update the Project 4 tracking flow around the issue-presence
check to add the assigned issue when missing, then set its status to “In PR
Review.”
- Around line 84-109: Update the delivery workflow around the convention review
and push steps to require committing each repair, then rerunning format:check,
lint, typecheck, tests, and the code-reviewer before pushing or completing the
task. Ensure the same gate applies to the referenced flow at lines 222–225, and
do not allow pushing uncommitted repairs.
In @.claude/commands/dep-bump.md:
- Around line 39-46: Update the validation instructions in the dep-bump command
so that after running npx expo install --fix for SDK patch drift, rerun both npx
expo install --check and npm run doctor, preserving the existing blocking
validation flow.
In @.claude/commands/generate-pr-description.md:
- Around line 3-7: Update the overview for the generate-pr-description command
to state that it is supported in both Claude Code and Cursor, while preserving
the existing text-only behavior and template reference.
- Around line 112-121: Align both workflow surfaces with the canonical CI gate:
in .claude/commands/generate-pr-description.md lines 112-121, replace npx tsc
--noEmit with npm run typecheck; in .claude/commands/handle-dependabot.md lines
23-25, require format:check, lint, typecheck, npm test -- --ci, Expo Doctor when
applicable, and all required build checks before approval.
- Around line 76-80: Update the PR title example in the relevant template
section to use the mandatory `[`#38`]:` format instead of `#38:`; keep the branch
name and suggested type unchanged.
- Around line 24-31: Update the title-generation guidance in the Format section
to require a GitHub issue number for every change. Remove the “No ticket”
descriptive-prefix fallback, and instruct the agent to stop and request an issue
number when no issue can be identified before generating a PR title.
In @.claude/commands/handle-dependabot.md:
- Line 20: Update the open-PR listing and subsequent triage/merge eligibility in
the Dependabot workflow to accept only the approved Dependabot author identities
and PRs whose baseRefName is "main". Strengthen the safe-PR gate to require
successful Lint & Format, Unit Tests, and Quality Gates checks, including
TypeScript, Expo Doctor, and expo install --check, before proceeding.
In @.claude/commands/onboard.md:
- Around line 11-19: Update the onboarding command sequence so Expo MCP
authentication via mcp_auth occurs before Step 1 and before any Expo-related
command, including npm run prebuild. Apply the same authentication gate before
npm run doctor and other Expo commands, while preserving the existing setup
steps and ordering where no Expo command is involved.
In @.claude/commands/start-issue.md:
- Around line 129-149: Update the GraphQL query used by the ITEM_ID assignment
to paginate issue projectItems using pageInfo endCursor and hasNextPage until
all Project 4 items are retrieved. Adjust the jq filter to select the Project 4
node and output only its scalar .id value, preserving the existing re-query flow
after project item-add.
In @.claude/skills/sync-claude-primitives/SKILL.md:
- Around line 42-57: Update the target list in the sync skill to include
.github/copilot-instructions.md alongside CLAUDE.md and AGENTS.md, and require
auditing it against the same agent contracts so cross-tool guidance remains
aligned.
- Around line 146-152: Clarify the Phase 5 workflow so validation runs before
updating .claude/.config-sync-state.json, and treat the state-file edit as an
explicit workflow-generated change excluded from the approved findings
comparison. Specify that lastSyncedSha records the source HEAD whose approved
changes passed validation, preventing failed runs from marking the working tree
as synced.
- Around line 150-153: Update the sanity-check step in the workflow instructions
to require npm run format:check, npm run lint, npm run typecheck, and npm test
-- --ci for every change, removing the conditional reference to scripts or
config files before chaining to /create-pr.
In `@docs/AGENT_ONBOARDING.md`:
- Line 244: The related-docs command index in the onboarding documentation is
missing /generate-pr-description. Add this command to the list alongside the
existing slash commands, preserving the current alias and canonical/shim
references.
In `@docs/issue-tracking.md`:
- Line 67: Update the PR title guidance in the issue-tracking documentation to
remove the alternate “#NNN: Short description” format, leaving only the
canonical “[`#NNN`]: Short description” format required by the repository
workflow.
In `@scripts/architecture-guard.mjs`:
- Around line 10-20: The allow function should emit the Claude PreToolUse
response schema: move the permission decision into hookSpecificOutput as
hookEventName and permissionDecision: "allow", while retaining agentMessage as
additionalContext there so the Cursor adapter still receives it. Remove the
current top-level permission and agent_message fields, then preserve the
existing process exit behavior.
---
Nitpick comments:
In @.claude/skills/write-test/SKILL.md:
- Around line 33-37: Update the service-testing rules in the skill document to
explicitly require mocking global fetch for endpoint unit tests and restoring
the mock after each test, following neighboring service *.test.ts patterns. Keep
the existing per-file jest.mock guidance and live API opt-in instructions
unchanged.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 7e07ec71-aa61-4a3c-a1b9-c7c742b662a5
📒 Files selected for processing (33)
.claude/agents/code-reviewer.md.claude/commands/create-pr-branch.md.claude/commands/create-pr.md.claude/commands/dep-bump.md.claude/commands/generate-pr-description.md.claude/commands/handle-dependabot.md.claude/commands/onboard.md.claude/commands/open-pr.md.claude/commands/start-issue.md.claude/settings.json.claude/skills/sync-claude-primitives/SKILL.md.claude/skills/write-test/SKILL.md.cursor/commands/create-pr-branch.md.cursor/commands/create-pr.md.cursor/commands/dep-bump.md.cursor/commands/generate-pr-description.md.cursor/commands/handle-dependabot.md.cursor/commands/onboard.md.cursor/commands/open-pr.md.cursor/commands/start-issue.md.cursor/hooks.json.cursor/hooks/architecture-guard.mjs.cursor/rules/ai-tooling-bridge.mdc.cursor/rules/commands.mdc.cursor/rules/delivery.mdc.cursor/rules/docs.mdc.cursor/rules/pr-delivery-ci-gate.mdc.cursor/rules/rule-precedence.mdcAGENTS.mdCLAUDE.mddocs/AGENT_ONBOARDING.mddocs/issue-tracking.mdscripts/architecture-guard.mjs
Require ticketed PRs, committed repairs before push, and Claude-compatible hook output so agents cannot skip Project 4 or CI-parity checks.
JonathanSeehagen
approved these changes
Aug 13, 2026
JonathanSeehagen
left a comment
JonathanSeehagen
left a comment
Contributor
There was a problem hiding this comment.
Human + agent review
Validated against #324 acceptance criteria on branch mrace/chore/324-agent-delivery-tooling:
- Canonical commands under
.claude/commands/; Cursor shims are thin pointers only /start-issue→/create-prloop documented inAGENT_ONBOARDING.md,issue-tracking.md,commands.mdc- Local CI parity:
format:check,lint,typecheck,npm test -- --ci— all pass - Remote CI green on head SHA
- No app runtime changes; Android device QA N/A
Approve for merge.
Follow-up suggestion (non-blocking)
Today the agent surface is intentionally distributed (rules in .cursor/rules/*.mdc, commands in .claude/commands/, hooks in .cursor/hooks.json + docs.mdc). That works, but onboarding requires jumping across several files.
Consider a thin docs/agent-tooling-index.md (or similar) that centralizes links + one-line purpose for each rule, slash command, hook, agent, and skill — without duplicating full procedure text. Single map; canonical content stays in place. Happy to pick this up as a small chore issue if the team wants it.
JonathanSeehagen
approved these changes
Aug 13, 2026
mattrace-buildmidwestern
deleted the
mrace/chore/324-agent-delivery-tooling
branch
August 13, 2026 14:53
9 of 14 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
TLDR
Adds a shared agent command surface so Cursor and Claude Code follow the same issue-to-PR loop. Canonical instructions live under
.claude/commands/; Cursor/files are thin shims./start-issueclaims a Project 4 ticket, branches frommain, and chains to/create-pr, which now runs the convention reviewer, moves the card to In PR Review, and watches CI. Existing six commands keep working; Android-only,Refs #NNN, and npm gates are unchanged.Reviewer checklist
Refs #NNN— do not useCloses/Fixes/Resolves)AGENTS.md)Details
Refs #324
Agent delivery was split across Cursor-only slash commands with no single issue → branch → implement → PR path. This PR moves command bodies to
.claude/, adds/start-issueand an/open-pralias, and documents the loop in onboarding and issue-tracking. No app runtime behavior changes. Device QA is N/A (docs/hooks/commands only).Type of change:
Technical changes
.claude/commands/— canonical/start-issue,/create-pr,/create-pr-branch,/open-pralias, plus moved onboard/dep-bump/handle-dependabot/generate-pr-description.cursor/commands/— thin shims; new/start-issueand/open-pr.claude/agents/code-reviewer.md— Fluent layer/Android/theme/Refschecklist.claude/skills/write-test/andsync-claude-primitives/.cursor/rules/ai-tooling-bridge.mdcandpr-delivery-ci-gate.mdcscripts/architecture-guard.mjs— warn-onlyfetch(/SQL insrc/app/andsrc/components/docs/issue-tracking.md,docs/AGENT_ONBOARDING.md,CLAUDE.md,AGENTS.mdTesting
npm run format:check— passnpm run lint— pass (0 errors; pre-existing warnings in unrelated files)npm run typecheck— passnpm test -- --ci— 112 passed, 1 skipped (723 tests)How to verify
/palette lists/start-issue,/open-pr, and the existing six commands..claude/commands/start-issue.mdand.claude/commands/create-pr.md— facts should be Fluent only: basemain,Refs #NNN, Project 4 In Progress (Dev) / In PR Review, npm, Android-only (no iOS how-to)..cursor/commands/*.mdare shims that point at.claude/commands/with no duplicated procedure.npm run format:check && npm run lint && npm run typecheck && npm test -- --ci.Expected: Commands and docs describe one ticketed loop; CI stays green; no Product-column board moves; no closing keywords.
Follow-ups
None.
Summary by CodeRabbit
New Features
Documentation
Chores