The Agent Skills Directory

[COMMAND_EXECUTION]: The repomix_batch.py script executes the repomix CLI and npx via subprocess.run. It uses a list-based argument structure to prevent shell injection.
[EXTERNAL_DOWNLOADS]: The skill uses npx to fetch and run the repomix package for remote repositories. Repomix is a recognized community tool for preparing codebase context.
[PROMPT_INJECTION]: The skill's primary function is to ingest repository content for LLM context, which constitutes an indirect prompt injection surface.
Ingestion points: Reads files from repositories via the repomix CLI.
Boundary markers: Uses XML or Markdown markers for file separation; however, these are not robust against adversarial content.
Capability inventory: Subprocess execution and local file system writes.
Sanitization: Relies on repomix's internal Secretlint for credential detection but does not filter content for injection patterns.
[CREDENTIALS_UNSAFE]: The EnvLoader class in repomix_batch.py searches for and loads environment variables from several parent directory locations, including .env files that may contain sensitive agent configuration.

repomix