YourAI.tools
Red Hat Developer logo

Skill

rhdh-pr-review

review pull requests for RHDH

Published by Red Hat Developer Updated Jul 16
Covers GitHub Code Review QA Engineering

Description

Review pull requests: code-level analysis with inline comments, and live cluster testing for rhdh-operator PRs. Supports GitHub (GitLab planned). Use when asked to review a PR, review code, post review comments, test PR changes on a cluster, deploy PR images for testing, or do a full PR review. Also use when given a PR URL or number and asked for feedback, or when user mentions "review this PR", "PR review", "code review", or "test this PR on my cluster".

SKILL.md

<cli_setup>

For cluster testing workflows, set up the orchestrator CLI:

RHDH=../rhdh/scripts/rhdh

</cli_setup>

<essential_principles>

Reviews follow a three-layer pipeline: **fetch** (forge-specific) → **analyze** (agnostic) → **post** (forge-specific). Each layer produces a structured artifact for the next. The analyze layer never calls forge-specific CLIs. Reviewers will produce false positives. Verify every finding against actual code at HEAD before including it. Drop findings that reference non-existent code, duplicate existing comments, misread the code, or conflict with codebase conventions. Present the full review draft — summary, inline comments with file:line, and event type — to the user before posting. Proceed only after confirmation. For cluster testing: deploy the full PR bundle/manifests, not just the operator binary image. PR changes to CRDs, RBAC, default config, or bundle metadata are baked into the OLM bundle or install.yaml — a binary-only image swap misses them.

</essential_principles>

What would you like to do?

Code Review

  1. Review PR code — Analyze a PR diff, generate findings, and post inline comments
  2. Review PR code (analysis only) — Analyze without posting (e.g., to review locally or post later)

Cluster Testing (rhdh-operator PRs)

  1. Test PR on cluster — Deploy PR operator bundle on a live RHDH cluster and verify changes

Combined

  1. Full review — Code review + post to GitHub + cluster testing

Wait for response before proceeding.

ResponseWorkflow
1, "review", "review PR", "code review", a PR URL or numberworkflows/fetch-github.md → workflows/review-code.md → workflows/post-to-github.md
2, "analyze", "analysis only", "review locally"workflows/fetch-github.md → workflows/review-code.md (stop after findings)
3, "test", "cluster", "deploy", "operator PR", "test on cluster"workflows/fetch-github.md → workflows/review-operator-pr.md
4, "full", "full review", "both"workflows/fetch-github.md → workflows/review-code.md → workflows/post-to-github.md → workflows/review-operator-pr.md

Routing rules

  1. PR URL or number with no other context: default to route 1 (code review + post).
  2. rhdh-operator PR detected (repo is redhat-developer/rhdh-operator): suggest route 4 (full review) but let the user choose.
  3. "review" without "post" or "cluster": route 1.
  4. All routes start with fetch. The fetch workflow produces a context artifact consumed by all downstream workflows.

Forge detection

Currently GitHub only. Detect from URL pattern or gh CLI availability.

When GitLab support is added: fetch-gitlab.md and post-to-gitlab.md will slot into the same pipeline. The analyze workflow (review-code.md) is forge-agnostic and needs no changes.

<artifact_contracts>

Context artifact (fetch → analyze / cluster test)

Produced by the fetch workflow, consumed by all downstream workflows:

context artifact
├── forge: "github"
├── repo: "owner/repo"
├── pr_number: 123
├── head_sha: "abc123..."
├── base_ref, head_ref, title, body, author, state, url
├── labels: [...]
├── files: [{path, additions, deletions}, ...]
├── total_additions, total_deletions
├── diff: "full unified diff"
├── linked_issues: [{number, title, body, labels, state}, ...]
├── jira_keys: ["RHIDP-1234", ...]
├── existing_comments: [{user, path, line, body, created_at}, ...]
├── existing_reviews: [{user, state, body}, ...]
└── ci_status: "pass" | "fail" | "pending" | "unknown"

Findings artifact (analyze → post)

Produced by the analysis workflow, consumed by the posting workflow:

findings artifact
├── pr: {repo, number, head_sha}
├── summary: "top-level review text"
├── event: "COMMENT" | "APPROVE" | "REQUEST_CHANGES"
└── findings[]
    ├── path, line, start_line
    ├── type: "suggestion" | "question" | "observation"
    └── body: "comment text"

</artifact_contracts>

<reference_index>

ReferencePurposeLoad when...Path
review-perspectivesReview perspective examples and signal hintsRunning review-code.mdreferences/review-perspectives.md
operator-pr-imagesCI image extraction and validationRunning review-operator-pr.mdreferences/operator-pr-images.md
github-referencegh CLI patterns, PR queriesRunning any GitHub workflow../rhdh/references/github-reference.md (if available)
rhdh-reposRHDH ecosystem repository mapCluster testing../rhdh/references/rhdh-repos.md (if available)

</reference_index>

<skills_index>

SkillPurposePath
rhdhOrchestrator, environment status, activity tracking../rhdh/SKILL.md

</skills_index>

<success_criteria>

Code review

  • PR context fetched (metadata, diff, linked issues, existing comments)
  • Review perspectives chosen based on PR content
  • Findings verified against actual code at HEAD
  • False positives dropped with reasoning shown
  • Review draft presented to user with event type choice
  • Review posted to forge (if posting route selected)

Cluster testing

See workflows/review-operator-pr.md <success_criteria> for the full checklist.

Full review

All code review criteria + all cluster testing criteria.

</success_criteria>

AboutPrivacy PolicyTerms and Conditions

© 2026 YourAI.tools. Every skill from an identity-verified publisher.

Independent catalog. Not affiliated with, endorsed by, or sponsored by Anthropic or any listed publisher. All trademarks belong to their respective owners.