Savage Roast AI (No Mercy)

• Verdict: Pass
• Confidence: High
• Analyzed: Apr 1, 2026, 6:36 AM
• Findings: 0
• Version: 1.0.1

Overview

The skill's requested resources, files, and runtime instructions are coherent with its stated purpose (generating one-line internet-style roasts); no hidden endpoints, extra credentials, or risky install steps were found — though examples and docs have small inconsistencies and the content is intentionally provocative so exercise caution when using it on real people.

This skill is internally coherent for producing one-line internet-style 'roasts' and does not request credentials or install external code. Before installing, consider: - Content safety: the skill's purpose is to produce provocative text; although prompts forbid abusive/discriminatory content, enforcement depends on the underlying LLM — test thoroughly to ensure it won't produce harassment or violate platform policy. - Inconsistency: SKILL.md enforces a single-sentence output but examples include multi-sentence roasts; clarify which behavior will be enforced. - Privacy: user-provided inputs will be transformed into shareable roasts — avoid pasting private or identifying information if you plan to share outputs. - Implementation: the script currently returns canned demo responses and contains a comment about calling an LLM API in production; confirm which LLM/OpenClaw capability will actually be used and whether any model endpoints will receive user text. If you need stricter guarantees (e.g., to prevent harassment or to audit where inputs are sent), ask the skill author to: (1) make the output-length enforcement consistent with examples, (2) document the exact runtime LLM call and any content-moderation safeguards, and (3) confirm no external endpoints are contacted.