The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's primary function is to download and execute code from remote GitHub repositories. It automates the cloning process and runs project-specific entry points like npm start, python main.py, cargo run, and go run.
[COMMAND_EXECUTION]: The skill performs automated shell command execution for environment setup and project execution. This includes installing system dependencies, creating virtual environments, and running containerized services via docker-compose.
[DATA_EXFILTRATION]: The skill is instructed to find and configure sensitive files such as .env and config.yaml. It explicitly asks for API keys and database credentials if they are missing, which could lead to the exposure of sensitive user data if the project being replicated contains malicious logic.
[PROMPT_INJECTION]: The skill exhibits a high vulnerability to indirect prompt injection (Category 8) because it is designed to ingest and follow instructions found in external, untrusted project documentation.
Ingestion points: Phase 2 involves reading and analyzing README.md files and project source code to determine setup and execution steps.
Boundary markers: There are no boundary markers or instructions to disregard malicious commands embedded within the documentation of the target repository.
Capability inventory: The skill possesses extensive capabilities including directory creation (mkdir), file deletion (rm), package installation (npm install, pip install), and arbitrary process execution.
Sanitization: No sanitization or validation of the logic extracted from external projects is implemented before execution.

github-project-replication