The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides explicit instructions to clone an external repository (git clone https://github.com/Champbreed/AegisOps-AI.git) and execute a Python script (python3 main.py) from that repository. This pattern of downloading and immediately running code from an untrusted third-party source represents a high risk for arbitrary code execution on the host machine.
[EXTERNAL_DOWNLOADS]: The skill directs the agent to download external resources from an untrusted GitHub account (Champbreed) that does not match the verified author context. Additionally, it requires the installation of external Python packages (google-genai, python-dotenv).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data, including Git diffs for Linux kernel patches and Terraform cost analysis reports.
Ingestion points: Raw Git diffs (Linux Kernel patches) and infrastructure-as-code (Terraform plan) outputs are ingested into the agent context for analysis.
Boundary markers: The instructions do not define boundary markers (such as XML tags or specific delimiters) to isolate this untrusted data from the agent's instructions.
Capability inventory: The skill possesses the capability to execute shell commands (git, python3, pip) and make network calls via the Google GenAI SDK.
Sanitization: There is no evidence of input sanitization or validation mechanisms to prevent the agent from interpreting instructions embedded within the patches or plans as its own high-level commands.
[COMMAND_EXECUTION]: The skill includes instructions to execute several local shell commands, including environment setup (python3 -m venv), package management (pip install), and direct script execution (python3 main.py), which are performed on content retrieved from an external source.

aegisops-ai