brutal-honesty-review

The Brutal Honesty Review skill is an advanced quality engineering diagnostic tool designed for high-stakes technical environments where ambiguity and sugar-coating lead to architectural debt or critical failures. By channeling the archetypes of Linus Torvalds for technical precision, Gordon Ramsay for quality standards, and James Bach for adversarial BS-detection, this agent provides actionable, high-frequency feedback that strips away vendor hype and sloppy coding practices. It is intended for senior engineers and architect-level contributors who require a direct, unfiltered assessment of their work rather than polite encouragement. Typical use cases include evaluating complex architectural decisions, auditing security-sensitive code paths, identifying gaps in test suites, and challenging industry 'best practices' that lack empirical validation.

• Employs three distinct review modes: Linus (technical fundamentals and correctness), Ramsay (standards and quality excellence), and Bach (critical BS-detection and risk assessment).

• Enforces a minimum of 3 weighted findings (Critical, High, Medium, Low) to ensure that the reviewer remains thorough and evidence-based in every session.

• Provides a structured output format—Broken, Wrong, Correct, Fix, and Impact—ensuring that every critique includes an actionable path forward and a clear definition of excellence.

• Calibrates tone dynamically through intensity levels (Direct, Harsh, Brutal), allowing users to modulate the feedback intensity based on the severity of the technical issues.

• Integrates with coding agents to analyze PR diffs, vendor claims, and test coverage metrics, serving as a high-velocity quality gate for CI/CD pipelines.

• Designed for environments with high psychological safety where the focus is on the work, not the worker; do not use this tool on junior developers' first contributions or with demoralized teams.

• Requires explicit context, such as a PR diff, a set of architectural requirements, or a list of vendor claims as input to generate a valid assessment.

• Operates best when the user explicitly defines the desired mode and calibration level to match the severity of the problem.

• Acts as a final validation step in a broader automated testing suite, ensuring that the human-machine collaboration maintains the highest standards of software integrity and engineering rigor.