jira-rework

name	jira-rework
description	Researches a Jira issue sent back for rework: reads the issue, comments, and codebase to understand QA/UX feedback, then proposes fix options. Triggers: rework, rework issue, QA feedback, QA issue, investigate issue, research issue, UX feedback from Jira.
argument-hint	<issue-key> e.g. PROJ-1234 [optional context]
metadata	{"boost-tags":"jira github","schema-required":"^1"}

Jira Rework Research

Research a Jira issue that has been sent back for rework — understand the feedback, investigate the codebase, and propose options for fixing it.

Jira MCP tool calls in this skill use your project's configured Jira MCP server namespace. The tool references below are written mcp__<jira>__jira_* (or a bare jira_* for brevity); substitute <jira> with the namespace . The actual call is always the fully-qualified mcp__<jira>__jira_* — e.g. with the namespace mcp-atlassian, mcp__mcp-atlassian__jira_get_issue. A bare jira_* name is NOT a callable tool.

Project scope: the configured project key is ; cross-project refusal is . When refusal is true and the issue key's prefix differs from the configured key, refuse the rework research and tell the user.

When to Use This Skill

Use this skill when:

• A Jira issue has been moved to a "Rework" status after QA or UX review
• QA has reported a problem that needs investigation before fixing
• A design/UX review has flagged problems with an implementation

Do NOT use for:

• Implementing a feature from scratch (use implement-spec)
• Applying PR review feedback before merge (use pr-review-feedback)
• General bug reports not tied to a Jira rework (use bug-fixing)

Workflow

Phase 1: Gather Context from Jira

1. Fetch the Jira issue:

   mcp__<jira>__jira_get_issue(issue_key: "<ISSUE-KEY>")
   

   Extract the summary and description (the original requirement), the current status and assignee, and any linked issues or parent epic.

2. Read all comments to find QA/UX feedback, requested changes, screenshots or reproduction steps, and clarifying discussion. Focus on the most recent comments — later ones often override earlier feedback.

3. Summarize the feedback into: the original requirement, what was implemented, what's wrong, and who raised it (QA, UX, PO — this affects the kind of fix needed).

Phase 2: Find the Implementation

1. Find the PR for this issue:

   gh pr list --search "<ISSUE-KEY> in:title" --state all --json number,url,title,state
   

2. Get state, branch, and file list:

   gh pr view <number> --json state,headRefName,mergedAt,merged,files,url
   

   Use gh pr diff <number> separately if you need to read the changed code — after narrowing which files matter.

3. Determine the PR state — it affects the fix workflow:

   • Merged (merged: true): the code is already integrated. The fix needs a new branch off the base branch the PR targeted.
   • Open (merged: false, state: "OPEN"): the fix can be pushed to the existing branch.
   • Closed, unmerged (merged: false, state: "CLOSED"): the previous PR was abandoned — treat as not merged and start a fresh fix branch.

4. Read the relevant code — focus on the files most likely related to the reported issue.

Phase 3: Investigate the Problem

Investigate using whatever inspection tools the project's stack provides — a REPL / console, schema inspection, application logs, browser or console logs for UI issues. Based on the feedback type:

• QA bug / incorrect behaviour — read the code paths for the reported scenario; look for logic errors, missing edge cases, or wrong conditions; check whether the test suite covers the failing scenario.
• UX / design issue — read the view / UI code; compare the implementation against the feedback; check for missing states (loading, empty, error); look at similar features for patterns that should have been followed.
• Missing functionality — compare the original requirement against what was implemented; flag whether the missing piece was part of the original spec or is a new requirement.

Phase 4: Propose Options

Present findings and options to the developer:

## Rework Summary: <ISSUE-KEY>

### Original Requirement
[What the feature was supposed to do]

### Feedback
[Concise summary of what QA/UX reported, with attribution]

### Root Cause
[What's actually wrong — code-level explanation with file:line references]

### Options

**Option 1: [Short description]**
- What: [the approach]
- Files: [files that would change]
- Effort: Small / Medium / Large
- Trade-offs: [downsides or considerations]

**Option 2: [Short description]**
- What: [the approach]
- Files: [files that would change]
- Effort: Small / Medium / Large
- Trade-offs: [downsides or considerations]

### Recommendation
[Which option you recommend and why]

For a straightforward fix with one obvious solution, a single recommendation is enough — don't force multiple options.

After the Developer Chooses

1. Create a fix branch from the correct base:
   • If the original PR is merged, branch from the same base branch it targeted (rework is typically found on a staging / integration environment).
   • If the original PR is open, switch to its branch and push the fix there.
2. Implement the fix following the chosen approach.
3. Update or add tests covering the rework scenario.
4. Run the project's code-style and quality checks (see the backend-quality / frontend-quality skills for the relevant stack).
5. Commit and push.
6. If the original PR was merged, open a new PR referencing the Jira issue.
7. Update Jira if the fix changes user-facing behaviour (use the jira-updates skill).
8. Transition the issue to the appropriate review status — discover the transition via jira_get_transitions; never hardcode transition IDs.

Guidelines

• Don't start fixing immediately — research first, propose options, let the developer decide.
• Read the full comment thread — later comments often clarify or override earlier feedback.
• Consider the feedback source — QA issues usually need exact behaviour fixes; UX issues may need design discussion first.
• Check if it's a bug or a requirement change — rework feedback sometimes adds new requirements that weren't in the original spec. Flag this clearly.
• Look at the test suite — if the current tests pass but the behaviour is wrong, the tests may need updating too.
• Be specific about file locations — reference exact files and line numbers when discussing the root cause.