arch-review

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill instructs the agent (and a subagent) to read and include full diffs and complete file contents and to cite specific file:line snippets, so any secrets present in those files would be passed through the LLM and could be output verbatim, creating an exfiltration risk.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly calls gh pr diff <number> / accepts a GitHub PR URL (e.g., https://github.com///pull/) at runtime and injects the fetched diff and full file contents into the subagent prompt, so remote GitHub PR content can directly control the agent's prompts.