The Agent Skills Directory

[Data Exposure & Exfiltration] (LOW): Accesses ~/.skillforge-config to retrieve paths for the marketplace. While accessing files in the home directory is a sensitive operation, it is required for the skill's configuration, and no evidence of data exfiltration to external domains was found.
[Indirect Prompt Injection] (LOW): The skill exhibits a vulnerability surface where untrusted data (plugin content, user-provided names, and configuration values) is interpolated into prompts sent to subagents.
Ingestion points: ~/.skillforge-config, .claude/ directory contents, and interactive user input for plugin names and versions.
Boundary markers: Absent. The skill uses direct string interpolation (e.g., prompt="发布插件 {plugin-name} 到 {marketplace-path}") without delimiters or instructions to ignore embedded commands.
Capability inventory: Uses Bash for file system manipulation, Write for metadata updates, and Task for triggering external agent workflows.
Sanitization: None detected. The skill does not appear to validate or escape inputs before passing them to the Task tool.
[Command Execution] (LOW): The skill utilizes the Bash tool to perform standard directory management and file copying operations (mkdir, cp). These actions are limited to the workspace and development directories.

sync-to-marketplace