The Agent Skills Directory

External Downloads (LOW): The skill requires the installation of the repomix CLI tool via npm install -g repomix or brew install repomix. As this tool is not from a source on the pre-approved trusted list, it is considered an unverifiable dependency.
External Downloads (LOW): The skill facilitates downloading code from arbitrary remote repositories using the --remote flag. While this is the core functionality, it involves network operations to non-whitelisted domains to fetch third-party content.
Indirect Prompt Injection (LOW): By packaging entire repositories for LLM analysis, the skill creates a surface where an attacker-controlled repository could contain malicious instructions hidden in code or comments.
Ingestion points: Local and remote repository files are read and aggregated via the repomix command in SKILL.md and repomix_batch.py.
Boundary markers: The tool implements XML and Markdown separators to delimit files, which provides structural context but does not fully prevent instruction override in adversarial scenarios.
Capability inventory: The skill executes the repomix binary via shell commands (subprocess context).
Sanitization: The tool uses Secretlint to detect hardcoded credentials (API keys, secrets), but does not perform sanitization of natural language instructions within the processed data.

repomix