review-package

[Skill Scanner] Instruction to copy/paste content into terminal detected The provided workflow description is coherent with its stated purpose of building a review package for external analysis. It is not inherently malicious and does not request credentials or perform network communication. However, it relies on local, potentially untrusted components (the packaging script and the analyzer agent) and uses clipboard for prompt transfer, which could pose misuse risks in shared environments. The approach is benign in intent but warrants caution about environment integrity and the trustworthiness of local tooling. LLM verification: This skill is functionally coherent with its stated purpose, but it contains a significant supply-chain/execution risk: it finds and executes a create-review-zip.sh under ~/.claude without verifying its integrity or requiring an explicit trusted path. That allows arbitrary local code execution with access to project files and temp file paths, which could be abused for exfiltration or other malicious actions. There are no remote URLs or embedded malware in the skill text itself, so direct malicio