How does 'Harsh Critique' differ from a standard AI feedback prompt?

This skill solves the problem of 'toxic positivity' in AI feedback by enforcing a strict scoring rubric, limiting praise, and requiring specific remediation steps for every technical flaw identified.

Which AI agents or models are compatible with this skill?

The skill is framework-agnostic and compatible with any text-based LLM or AI agent capable of processing technical documentation, code, and structured prompts.

What specific components are included in this purchase?

You receive the complete logic package for the Harsh Critique skill, including the scoring framework, severity classification guidelines, and decision logic for rebuild triggers.

Do I need to provide specific criteria for the skill to generate a score?

Yes, for the scoring and severity classification to function effectively, you must provide the artifact draft along with the specific success criteria or technical requirements you want it measured against.

Will this skill automatically rewrite my code or documentation once it identifies an error?

The skill is designed to identify and categorize technical debt and blockers; however, it focuses on audit and remediation instructions rather than performing the actual rewrite of your code or text.

How much configuration is required before I can use this for a code audit?

Setup is immediate; once added to your workspace, you simply input your draft and requirements to begin the high-standard evaluation loop.

Harsh Critique for Digital Products (Rebuild Gate)

HARSH CRITIQUE — DRAFT
Product: Local Cache Function
Round: 1

SCORES

Thread safety — 1/10 — The global dictionary is not protected by locks, leading to race conditions in multi-threaded environments.
Memory efficiency — 2/10 — No TTL or eviction policy (LRU) is implemented; the dictionary will grow indefinitely until OOM.
O(1) lookup — 9/10 — Uses native dict hashing which provides average O(1) time complexity.

OVERALL: 4/10 (Formula: [1+2+9]/3)

ISSUES
BLOCKER: Lack of thread synchronization. Fix: Implement threading.Lock around set and get operations.
BLOCKER: No memory bounds. Fix: Implement a maximum size and use collections.OrderedDict to pop the oldest item when the limit is reached.
MAJOR: Global state pollution. Fix: Encapsulate the cache in a class instance rather than a global variable.

WHAT WORKS (max 3)

Lookup speed because it utilizes Python's optimized dictionary implementation.

RECOMMENDATION: REBUILD

HANDOFF
Return to Builder. The current implementation is unsafe for production use due to lack of concurrency controls and memory management.

STATUS: DRAFT ONLY — not live

About this skill

Scores any digital-product draft against clear success criteria. Every issue gets a concrete fix ranked BLOCKER / MAJOR / MINOR. Recommends REBUILD or ADVANCE TO TESTER. Built for Product Loop teams (Researcher → Builder → Critic → Tester). Draft and recommend only — never publishes or edits live shops.

Frequently Asked Questions

Harsh Critique for Digital Products (Rebuild Gate)