scaffold-rules
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- Remote Code Execution (HIGH): The skill invokes
bunx @plaited/development-skills, which downloads and executes code from the npm registry at runtime. - Evidence: Found in
SKILL.md:bunx @plaited/development-skills scaffold-rules. - Risk: The
@plaitedscope is not on the trusted organizations list. Executing unverified third-party code allows for arbitrary command execution on the host system. - Indirect Prompt Injection (LOW): The skill is designed to modify
AGENTS.mdandCLAUDE.md, which are used to provide instructions to AI models (e.g., Cursor or Claude). - Ingestion points: The skill writes to
AGENTS.mdand references it inCLAUDE.md. - Boundary markers: Uses
<!-- PLAITED-RULES-START -->and<!-- PLAITED-RULES-END -->to delimit content. - Capability inventory: The skill uses the
Bashtool to perform these modifications. - Sanitization: No explicit sanitization or validation of the rules content is visible in the skill definition, as the logic resides in the external package.
Recommendations
- AI detected serious security threats
Audit Metadata