Opus Review — Testing & QA agent for Claude Code
Read-only second-opinion reviewer for a finished spec or code diff — dispatched only when explicitly enabled via .sixth-sense/model-preferences.md, never for routine implementation or dispatch routing.
How to install Opus Review
Installs to ~/.claude/agents/dariovr1-sixth-sense-opus-review.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/dariovr1/sixth-sense/HEAD/agents/opus-review.md -o ~/.claude/agents/dariovr1-sixth-sense-opus-review.md Restart Claude Code, or start a new session, for it to be picked up.
What Opus Review does
name: opus-review description: Read-only second-opinion reviewer for a finished spec or code diff — dispatched only when explicitly enabled via .sixth-sense/model-preferences.md, never for routine implementation or dispatch routing. model: opus effort: high tools: Read, Bash, Grep, Glob maxTurns: 20
Opus Review
You are a second opinion, not a first pass. A cheaper model already reviewed this — triage/hunter for code, self-review/premortem for a spec. Don't repeat that work. You do no
Alternatives in Testing & QA
- Fable Verifier — Cold verifier for finished fable deliverables 837 ★
- Critical Thinking — Critical thinking mode that challenges assumptions and probes reasoning 100 ★
- Fixer Opus — Fix-only agent, tier 3. Diagnoses unknown-cause findings (reproduce, then fix) with a regression test, then co 77 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Router Sonnet
Worker subagent dispatched by the auto-model-router for standard development tasks — routine refactors, doc an
Verify Build
Runs the project's Gradle build/test for a change and reports a clean pass/fail with the relevant output. Use
Muse Rescue
Use as the second agentic lane, right after DeepSeek Harness, for bounded coding tasks such as a spec, rename,
Ada Devils Advocate
Use for a rigorous adversarial second opinion - stress-test a decision, design, or plan, poke holes, weigh tra
AI Provider Author
WRITE-access implementer for AI provider integrations, model routing, embeddings, prompt systems, streaming, a
Shadowcat Spec Reviewer Opus
Escalation twin of shadowcat-spec-reviewer — dispatch when shadowcat-spec-reviewer's findings read as shallow
Related Skills
Complete Task
Explicitly transition one identified work item to its tracker's completed state — a handler-dispatched, manual
Hq Done
Report a dispatched task finished, with the result in one line
Respond Reviewer
Classify a reviewer letter into per-comment labels and dispatch each to the right specialist; assemble a rebut