Unpinned, unverified remote content is treated as executable instructions
Source references: 2The Skill fetches a mutable `main/command.md` for every review, then says to apply “all rules” and follow its output-format instructions. The remote content is therefore treated as instructions rather than merely reference material. No visible restriction confines those instructions to interface guidance, and no commit pin or integrity check is specified.
If the remote file is changed by its maintainer, compromised, or supplied with malicious instructions, the reviewing agent could depart from the user's request, suppress or distort findings, or attempt other steps not authorized by the user. The evidence does not establish that such a change or attack has occurred.
The Skill explicitly fetches content from a mutable GitHub main branch before every review, treats it as containing all rules and output-format instructions, and applies every fetched rule. With no pinned revision or stated integrity check, that remote file can control the review and its response format. If it is compromised or changed inappropriately, findings could be manipulated, omitted, or pushed beyond the authorized UI review. Users can ask for an audited commit hash and a boundary that treats fetched material only as UI-rule data, never as higher-priority instructions.
1. Fetch the latest guidelines from the source URL below2. Read the specified files (or prompt user for files/pattern)3. Check against all rules in the fetched guidelines4. Output findings in the terse `file:line` formatShow 1 other places
Fetch fresh guidelines before each review:```https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md```Use WebFetch to retrieve the latest rules. The fetched content contains all the rules and output format instructions.