The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from session artifacts such as task.md and implementation_plan.md. * Ingestion points: Files located in the ~/.gemini/antigravity/brain/ subdirectory. * Boundary markers: None; the agent is instructed to read and summarize contents directly. * Capability inventory: Limited to file writing for report generation; no network or execution capabilities detected. * Sanitization: None.
[DATA_EXFILTRATION]: The skill reads from ~/.gemini/antigravity/brain/, a sensitive directory containing user session history. Access is necessary for the skill's primary function, and no network exfiltration patterns were detected.

analyze-project