The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute a local CLI utility (wewrite). It interpolates multiple variables directly into the command string, which is a classic command injection vector.
Evidence: In SKILL.md, the command wewrite publish {article} --cover {cover} --theme {theme} --title "{title}" --digest "{digest}" executes article publishing.
Risk: If the {title} or {digest} variables (which often come from the article content itself) contain shell metacharacters like backticks, semicolons, or dollar signs, an attacker could execute arbitrary code in the agent's environment.
[DATA_EXPOSURE]: The skill accesses article files and metadata through the Read and Glob tools to extract content for publishing. While these operations are functional, when combined with the command execution vulnerability, they create a path for exfiltrating sensitive data found in the environment if the article content is untrusted.

wewrite-publish