antigravity-marketing-lab
Fail
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill fetches and executes a shell script for installing the Antigravity CLI using a curl pipe to bash pattern. The script is hosted on an official domain belonging to a well-known service (Google).
- Evidence:
curl -fsSL https://antigravity.google/cli/install.sh | bash - [PERSISTENCE]: The skill instructs the user to modify their shell configuration file (
~/.bashrc) to ensure the CLI tool path persists across different sessions. - Evidence:
echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.bashrc - [INDIRECT_PROMPT_INJECTION]: The agent is designed to ingest and process content from a user-supplied brief file, creating an entry point for untrusted instructions.
- Ingestion points:
challenge/my-brief.md(read by agent skill) - Boundary markers: No explicit delimiter or protection instructions found in the skill workflow.
- Capability inventory: File system access for reading and writing marketing outputs using the
agyCLI. - Sanitization: No evidence of input validation or sanitization for the brief content.
- [EXTERNAL_DOWNLOADS]: The skill performs downloads from external sources, including a CLI installer and a marketing laboratory repository.
- Evidence:
curl -fsSL https://antigravity.google/cli/install.shandgit clone https://github.com/jarsing/antigravity-marketing-lab.git
Recommendations
- HIGH: Downloads and executes remote code from: https://antigravity.google/cli/install.sh - DO NOT USE without thorough review
Audit Metadata