The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and analyze potentially untrusted data from the Antigravity CLI's local storage. This creates a surface for indirect prompt injection if the files contain malicious instructions.
Ingestion points: The agent is directed to use read_file on log files (~/.gemini/antigravity-cli/log/cli-*.log), conversation histories (~/.gemini/antigravity-cli/conversations/), and brain artifacts.
Capability inventory: The agent has access to the terminal tool, which can execute shell commands.
Boundary markers: There are no explicit instructions or delimiters provided to ensure the agent ignores instructions embedded within the data it reads.
Sanitization: The skill does not implement validation or sanitization of the content found in logs or history files before processing.
[COMMAND_EXECUTION]: The skill is centered around executing shell commands via the terminal tool. It provides a comprehensive reference for the agy CLI's command surface, which includes the ability to run arbitrary shell commands through the CLI's internal prompt helpers (documented as !).
[EXTERNAL_DOWNLOADS]: The documentation within the skill (references/cli-docs.md) describes installation methods that fetch scripts from https://antigravity.google/ and pipe them directly into system shells (bash, PowerShell, or CMD). These resources originate from the official infrastructure of a well-known service.
[PRIVILEGE_ESCALATION]: The skill documents the use of the --dangerously-skip-permissions flag. This flag is designed to bypass the CLI's internal security prompts and permission checks during execution, which could be leveraged to perform sensitive actions without standard oversight.

antigravity-cli