publicity-review

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The reviewer is fed full diffs/files and is required to emit exact-match suggested_edits with an old_string that must match the offending lines, which forces the LLM to include secret values verbatim in its JSON verdict despite snippet redaction—creating an exfiltration risk.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The skill contains no explicit backdoor code but its design intentionally sends full diff/file contents to an external "Agent" and automatically applies agent-supplied edits without human review, which creates a high-risk capability for data exfiltration (leaking secrets), automated credential theft, and automated insertion of malicious edits/supply-chain backdoors if the Agent or its prompts are compromised or malicious.