dym-setup
Warn
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill constructs file paths at runtime using the user-provided
<slug>argument, such as.claude/skills/dym-setup-<slug>/SKILL.md. It then directs the agent to load and follow the instructions contained in the resulting file. Dynamically loading code or instructions from computed paths based on untrusted input is a known attack vector for loading unauthorized content. - [INDIRECT_PROMPT_INJECTION]: The skill accepts an untrusted
<slug>argument that directly influences the instructions the agent follows. - Ingestion points: The
<slug>argument read fromARGUMENTSinSKILL.md. - Boundary markers: No markers or constraints are defined to delimit the input.
- Capability inventory: The skill allows reading and following instructions from files located in
.claude/skills/,doyourmagic/setup/, anddoyourmagic-bundles/setup/. - Sanitization: The skill lacks validation or sanitization to prevent path traversal attacks or to ensure the slug refers to an intended workflow file.
Audit Metadata