Fix GitHub Issue

You are executing the /fix-issue workflow. Follow each step sequentially. Do NOT skip steps.

Step 1: Read the Issue

Run gh issue view $ARGUMENTS --json title,body,labels,milestone to fetch the issue details. Parse the title and body to understand what needs to be done. If $ARGUMENTS is empty or not a valid issue number, ask the user to provide one.

Step 2: Assign the Issue

Run gh issue edit $ARGUMENTS --add-assignee @me to assign the issue to yourself.

Step 3: Create a Feature Branch

1. •Run git checkout master && git pull to ensure you're on the latest master.
2. •Derive a short slug from the issue title: lowercase, replace spaces/special chars with hyphens, truncate to ~50 chars.
3. •Run git checkout -b feature/$ARGUMENTS-<slug> to create the branch.

Step 4: Implement the Fix

Study the issue description and the relevant code. Use the appropriate specialized agent (rust-developer for Rust code, java-spring-expert for Spring, nextjs-fullstack-expert for web, etc.) to implement the fix. Make the minimal changes needed to resolve the issue.

Lock files for test fixtures

When creating or updating test fixtures for Node.js-based projects (e.g., E2E fixtures with package.json), never write lock files manually. Instead, generate them using the appropriate package manager:

• •npm: Run npm install --package-lock-only in the fixture directory to generate package-lock.json
• •yarn (classic): Run yarn install --mode update-lockfile or yarn install in the fixture directory to generate yarn.lock
• •pnpm: Run pnpm install --lockfile-only in the fixture directory to generate pnpm-lock.yaml

This ensures lock files are valid and match the actual dependency resolution output.

Step 5: Run Tests

Run cargo nextest run --features cuda to execute all tests. If any tests fail, fix them before proceeding. Re-run until all tests pass.

Step 6: Run Clippy

Run cargo clippy --all-targets --features cuda -- -D warnings. If there are any warnings or errors, fix every single one. Re-run clippy until completely clean — zero warnings, zero errors.

Step 7: Run Formatter

Run cargo fmt --all to format all code.

Step 8: Commit

Stage all changed files and commit using conventional commit format:

feat(<scope>): <short description> (#$ARGUMENTS)

Where <scope> is the affected crate or module (e.g., detect, pipeline), and <short description> summarizes the change. Use fix instead of feat if the issue is a bug fix, refactor for refactoring, etc.

Step 9: Code Review

Launch the code-reviewer subagent (sonnet model) to review the diff between master and your branch:

• •Ask it to review the output of git diff master...HEAD
• •Wait for the review results

Step 10: Address Review Feedback

Fix all Critical and Important issues identified by the code reviewer. For Suggestions, use your judgment — apply them if they're quick wins, skip if they'd over-engineer the solution.

After fixing:

1. •Re-run cargo nextest run --features cuda — all tests must pass
2. •Re-run cargo clippy --all-targets --features cuda -- -D warnings — must be clean
3. •Re-run cargo fmt --all

If any changes were made, create a second commit:

refactor: address code review feedback for #$ARGUMENTS

Step 11: Push and Open PR

1. •Push the branch: git push -u origin feature/$ARGUMENTS-<slug>
2. •Create the PR using gh pr create with:
   • •Title matching the first commit message
   • •Body containing:
     • •## Summary section with 1-3 bullet points describing what was changed and why
     • •Closes #$ARGUMENTS to auto-close the issue
     • •## Test plan section with a checklist of how the changes were verified