/pr-review

description: Run a comprehensive PR review, post results, then independently assess each finding
argument-hint: <pr-number> [context]
allowed-tools: Bash, Read, Grep, Glob, Task, TaskCreate, TaskUpdate, Skill, AskUserQuestion

Review PR

You are running a comprehensive review of a pull request, posting the results, then independently assessing each finding to separate genuine issues from nitpicks and false positives.

**User input:** $ARGUMENTS

**Note:** This command performs best with an Opus-class model. On Sonnet or Haiku, results may be shallower.

Step 0: Parse input and create task list

The user's input typically contains:

• A **PR number** (required)
• Additional **context**, focus areas, or constraints (optional)

Example inputs:

• `108`
• `108 error handling and test coverage`
• `108 focus on the new API endpoints`

Extract the PR number. If context was provided, note it for Step 2. If the input is ambiguous, ask the user to clarify.

After parsing input, create the progress-tracking task list. Create a task for Step 0 and immediately mark it in progress. Then create tasks for each of the remaining 8 steps one at a time, in step order, all starting as pending. Task list display order matches creation order, so each task must be a separate sequential call -- do not batch multiple task creations in a single message. Store each returned task ID for later use -- do not assume IDs are sequential.

| Task | Subject | activeForm | |------|---------|------------| | Step 0 | Step 0: Parse input and create task list | Parsing input | | Step 1 | Step 1: Check out PR branch | Checking out PR branch | | Step 2 | Step 2: Delegate PR review to pr-review-toolkit | Delegating PR review | | Step 3 | Step 3: Post review comment | Posting review comment | | Step 4 | Step 4: Run independent assessment | Running independent assessment | | Step 5 | Step 5: Post assessment comment | Posting assessment | | Step 6 | Step 6: Present CLI summary | Presenting summary | | Step 7 | Step 7: Handle deferred items | Handling deferred items | | Step 8 | Step 8: Recommend next steps | Recommending next steps |

Mark Step 0 complete.

Step 1: Check out PR branch

Mark Step 1 in progress.

Ensure you are on the correct branch:

gh pr checkout <pr-number>
git pull

Mark Step 1 complete.

Step 2: Delegate PR review to pr-review-toolkit

Mark Step 2 in progress.

Mark Step 2 complete when all sub-tasks of the delegation below are completed.

Use the Skill tool to invoke `/pr-review-toolkit:review-pr`:

• Pass `"Review PR #<pr-number>"` as the invocation string.
• **Always** include this instruction in the Skill invocation: "IMPORTANT: Do not run review agents in the background. For parallel execution, launch them in a single message instead."
• **Always** include this instruction in the Skill invocation: "You are authorized to use review-relevant agents from any installed plugin, not just the agents bundled with pr-review-toolkit. When launching review agents in parallel, also include any domain-relevant agents from other installed plugins that would provide useful analysis for the PR content (e.g., plugin-dev:skill-reviewer when reviewing skill definitions, plugin-dev:plugin-validator when reviewing plugin code). Only include supplementary agents when they are relevant to the content being reviewed."
• **Always** include this instruction in the Skill invocation: "If the PR has a linked issue (look for issue references like 'Fixes #N', 'Closes #N', 'Resolves #N', 'Part of #N', 'Issue #N', or a bare '#N' in the PR description), include the `feature-completeness-checker` agent alongside the other review agents. This agent verifies that the PR fully implements the requirements from the linked issue's acceptance criteria and implementation plan. Do not launch this agent if no linked issue is detected."
• **Always** include this instruction in the Skill invocation: "Label each Critical and Important finding with a sequential F-prefixed identifier (F1, F2, F3, ...) numbered continuously across both sections. Label each Suggestion with a sequential S-prefixed identifier (S1, S2, S3, ...) using a separate counter. Use bold prefixes in the output (e.g., `**F1:** Missing null check`, `**S1:** Consider extracting helper`)."
• **Always** include this instruction in the Skill invocation: "IMPORTANT: Create at least one sub-task to track your progress. Use `"Step 2.M: <action>"` as the subject format where M is the sequence number. Use best judgment on granularity based on the review scope."
• If context was provided, append the following as the last item in the Skill invocation block: `> **User context:** <context>`

Do NOT attempt to fix any issues -- this session is for review only.

Mark Step 2 complete. Do not stop here -- continue to Step 3.

Step 3: Post review comment

Mark Step 3 in progress.

After the review completes, post the full review results as a reply comment on the PR:

gh pr comment <pr-number> --body "..."

The comment must include:

• `<!-- mach10-review -->` as the very first line of the comment body (this invisible HTML marker enables reliable identification in future sessions)
• The complete review findings (Critical, Important, Suggestions, Strengths), including any findings from supplementary agents merged into the appropriate severity categories with inline source attribution (e.g., "per plugin-dev:skill-reviewer")
• F/S identifiers on every finding -- Critical and Important findings use `F<n>` numbered sequentially across both sections, Suggestions use `S<n>` with a separate counter (e.g., `**F1:** ...`, `**F2:** ...`, `**S1:** ...`)
• Model attribution at the bottom -- identify yourself by your actual model name (e.g., "Reviewed by Claude Sonnet 4.6" if running on Sonnet)
• A note that this is an automated review

Format the comment as a well-structured markdown document that can serve as input to a future `/mach10:pr-review-fix` session.

Use F/S identifiers (e.g.