resonance-ops-second-opinion

Dispatches diffs and decision artifacts to an independent second model and reconciles the findings.

Updated Dec 1, 2025
One-click install
npx skills add https://github.com/manusco/wolfgang --skill resonance-ops-second-opinion-manusco
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: resonance-ops-second-opinion
Source: https://github.com/manusco/wolfgang/tree/main/.agents/skills/ops/second-opinion
Command: npx skills add https://github.com/manusco/wolfgang --skill resonance-ops-second-opinion-manusco

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? A single model reviewing its own work rationalizes the same blind spots, so risky changes and one-way decisions ship with undetected defects. This Skill runs an independent second-model review and reconciles both views so disagreements surface the bug or assumption one model missed. ## Core Features & Use Cases - Diff review mode: Sends a code diff or PR to a configured independent reviewer via .forge/second_opinion.py, then reconciles findings into agreements, second-only, and primary-only buckets ranked P0-P3. - Decision review mode: Reviews confirmed goal contracts, ADRs, and migration plans for assumptions, evidence gaps, tradeoffs, reversal conditions, and the strongest objection, without inventing file lines or severities. - Independence and safety gates: Refuses to claim independence when reviewer identity matches the primary, treats empty or unanswered reviews as incomplete, and blocks dispatch of secrets or oversized artifacts. - Use Case: Before merging a JWT refresh-token rotation change, dispatch the diff to a second model, verify each of its findings against the actual code, and ship only after P0/P1 contradictions are resolved by evidence. ## Quick Start Ask the assistant to get a second opinion on the current diff before merging, providing a short note on what the change is supposed to do.

Frequently Asked Questions about resonance-ops-second-opinion

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get a second opinion on a code diff before merging?▼

Run the second opinion flow with a short context note describing what the change does. It dispatches the diff to a configured independent reviewer via .forge/second_opinion.py, then reconciles both reviews into agreements and disagreements ranked P0-P3.

How do I review an ADR or decision before any code exists?▼

Use decision mode with the artifact, for example --mode decision --artifact decision.md. The review covers assumptions, evidence gaps, tradeoffs, reversal conditions, and the strongest objection, without inventing file lines or P0-P3 severities.

What happens if no second model command is configured?▼

The script prints a manual review prompt you can run in another model such as Codex or a local model, then paste the findings back for reconciliation. Until that prompt is answered and reconciled, the gate stays incomplete rather than passing silently.

Can the same model act as the independent reviewer?▼

No. Independence requires a different configured reviewer identity, set with --reviewer-id or RESONANCE_REVIEWER_ID. If the identity is missing or matches the primary reviewer, the review is not independent and cannot satisfy the gate.

What if the two models disagree on a critical finding?▼

Do not average the disagreement into a medium confidence note. Open the actual code path, verify the claim against evidence, and decide on what the code shows; if still unresolved, escalate with both views.

When should I not use a second-model review?▼

Skip it for routine reversible work, vague brainstorming, creative taste calls, and simulated councils. It is meant for concrete high-risk artifacts: risky diffs, final goal-loop gates, and one-way decisions.