athena-pr-reviewer

Automate code review by orchestrating parallel multi-LLM analysis that gathers PR diffs, Jira context, and CLAUDE.md guidelines to surface actionable feedback. This Skill runs up to 8 specialized reviewers simultaneously—including Claude, Gemini, and Codex—filtering hallucinations against actual code and prioritizing findings flagged by multiple agents.

Yes. This Skill coordinates parallel reviews from specialized Claude reviewers, Gemini, and Codex agents in a single pass, each analyzing code quality, tests, errors, and documentation. Consensus across reviewers elevates critical issues, and all findings are verified against actual diffs to eliminate false positives.

The Skill cross-gathers PR metadata, diffs, Jira tickets, epics, git blame, prior comments, and CLAUDE.md guidelines automatically. It assembles this context into structured artifacts (context.md, pr.json, jira.json) and feeds them to parallel reviewers, ensuring feedback is grounded in full project context.

Hallucination filtering verifies all AI-generated review findings against actual code to eliminate false positives. This ensures only accurate, actionable feedback reaches you, reducing wasted time on phantom issues and improving review reliability across large or complex pull requests.

Yes. This Skill is designed for complex PRs with many files. It dynamically selects specialized reviewers, gathers cross-source context in parallel, runs all agents simultaneously, and presents prioritized, verified feedback—turning comprehensive reviews that would take hours into minutes.

You need GitHub for PR management, Jira for ticket context, and the gh CLI for GitHub integration. The Skill also requires acli for Jira, rclone for data transfer, and access to Claude, Gemini, and Codex APIs for multi-LLM review orchestration.