automation-self-care
Warn
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: MEDIUMPERSISTENCECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PERSISTENCE]: The skill implements a 'Plugin Bridge' that directs the agent to install a plugin into
~/.gemini/config/plugins/automation-self-care/and modify its core configuration file (~/.gemini/config/config.json). This is explicitly designed to bypass context budget limits and ensure the skill's instructions are persistently loaded into the agent's environment. \n- [COMMAND_EXECUTION]: The skill executes multiple local scripts and CLI tools to manage its lifecycle and automation plans. It runsscripts/install_antigravity_plugin.pyfor setup, usesscripts/build_core_set.pyto construct and lint automation plans, and relies on an external CLI tool namedacc(CareCenter) for system-level maintenance. \n- [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection because it ingests data from external application states, logs, and telemetry to dynamically tune its own prompts. It lacks clear boundary markers to distinguish between ingested data and instructions, and there is no documented sanitization process for the content ingested from these external surfaces before it influences the 'Prompt-Qualitts-Tuning' logic.
Audit Metadata