create-plugin
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input gathered from the user via the
scripts/gather-plugin-info.cjsscript. This input (name, description, keywords) is used to construct file paths and shell commands without sanitization.\n - Ingestion points: The
name,description, andkeywordsfields returned by the interactive terminal scriptgather-plugin-info.cjsreferenced inSKILL.md.\n - Boundary markers: Absent. The skill instructions do not include markers to delimit or escape user-provided data when used in subsequent steps.\n
- Capability inventory: Subprocess calls (
git,node,claude), file-write operations, and file-read operations described inSKILL.md.\n - Sanitization: Absent. The
scripts/gather-plugin-info.cjsscript only performs whitespace trimming and basic string splitting.\n- [COMMAND_EXECUTION]: The skill uses user-provided strings to construct shell commands, specificallyclaude plugin validate <repo-root>/plugins/<name>. If the user-suppliednamecontains shell metacharacters (e.g., semicolons, backticks), it could lead to arbitrary command execution.
Audit Metadata