What do the different verdicts mean?

PROCEED means the plan is sound. REVISE indicates significant issues that should be addressed first. HALT indicates fundamental problems that prevent safe progression until resolved.

What are 'Volkswagen patterns' in testing?

These are tests that verify only surface-level patterns or keyword matches (like checking if output length > 0) without checking the actual logic or semantic structure, giving a false sense of security.

Does the Critic Agent fix the code itself?

No, the Critic focuses purely on identifying issues and naming assumptions. It provides a verdict and feedback, but actual implementation or revision is left to the primary planning or implementation agents.

When should I invoke the Critic Agent?

The Critic Agent is most effective when invoked after completing a plan, reaching a conclusion from an investigation, or diagnosing a problem, but before any significant changes are applied.

Critic Agent

Name: Critic Agent
Author: nicsuzor

bynicsuzor

0•

Seguridad y Pruebas

Provides skeptical, second-opinion reviews of implementation plans and conclusions to identify logical flaws and untested assumptions.

The Critic Agent acts as an independent auditor within the Claude Code environment, specifically designed to challenge the reasoning of other agents and catch what might have been missed during the planning phase. By applying a skeptical lens, it identifies overconfident claims, scope drift, and missing edge cases before a single line of code is executed. It is particularly effective for reviewing test suites, where it flags superficial 'Volkswagen' patterns—tests that pass without verifying actual behavior—ensuring that your project maintains high-quality, semantically sound logic and strategic clarity.

Características Principales

01Ensures project scope alignment and catches missing edge cases early in the cycle.

020 GitHub stars

03Performs semantic verification for test code to prevent superficial passes.

04Detects logical errors, non-sequiturs, and circular reasoning in plans.

05Provides structured verdicts (PROCEED, REVISE, HALT) for streamlined decision-making.

06Identifies hidden assumptions and overconfident claims lacking evidence.

Casos de Uso

01Reviewing unit and integration tests for shallow or misleading assertions.

02Auditing a complex architectural migration plan before execution begins.

03Validating diagnostic conclusions and root-cause analyses after troubleshooting a bug.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add nicsuzor/academicops critic

For use in Claude.ai and ChatGPT

Download Skill