antigravity-skills-manager
Warn
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches skill data and configuration files from the vendor's repository (github.com/rmyndharis/antigravity-skills) using Python's standard library.
- [COMMAND_EXECUTION]: Executes system commands via
python3 skills_cli.pyor thenpx @rmyndharis/antigravity-skillspackage to manage the local skill lifecycle. - [PERSISTENCE]: The
installcommand writes content to~/.gemini/antigravity/skills/, a sensitive directory where the agent automatically discovers and loads instructions for all future sessions. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: Downloads instruction sets (SKILL.md files) from an external repository into the agent's active configuration path.
- Boundary markers: The instructions include manual safeguards, requiring the agent to confirm the skill ID with the user and display search results before installation.
- Capability inventory: Performs file-system write operations and network requests (urllib.request) to retrieve and store agent instructions.
- Sanitization: Relies on user confirmation and explicit path reporting to the user to mitigate unauthorized or silent modifications.
Audit Metadata