The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands and scripts to perform project discovery and filesystem operations. Specifically, scripts/dev-workflow.cjs uses execFile to invoke git for branch and status information, which is a restricted and safe method of command execution.
[DATA_EXPOSURE]: The skill aggregates content from .dev/ and .dev-archive/ directories across multiple projects into a single location (~/.dev-wiki/). While this involves reading sensitive project structure, the operation is local, and no data is transmitted to external servers.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted markdown files from various projects. It proactively mitigates the risk of indirect prompt injection by including a dedicated 'UNTRUSTED INPUT' section in the instructions, explicitly telling the agent that scanned content is data to be catalogued and must never be interpreted as commands.
[DYNAMIC_EXECUTION]: The Node.js script includes the js-yaml library but has explicitly disabled the dangerous js/function type during its build process to prevent potential code execution through YAML frontmatter in scanned files.

dev-wiki