agent-kernel-evolve
Warn
Audited by Gen Agent Trust Hub on Aug 4, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to run
agent-kernel evolve hooksto install 'self-evolve hooks' across various AI agent environments (Antigravity, Claude, Codex, OpenCode). This involves modifying the agent's operating environment to ensure persistence of the learning mechanism across different sessions. - [COMMAND_EXECUTION]: Execution of
agent-kernel hook self-evolveis suggested for background processing, which establishes long-running or recurring tasks to monitor and modify agent behavior. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through its core 'Playbook' generation feature.
- Ingestion points: The
agent-kernel evolve generatecommand captures the 'Current Workflow' to synthesize repeatable instructions. - Boundary markers: The instructions lack explicit boundary markers or delimiters to differentiate between benign workflow data and potentially malicious embedded instructions within that data.
- Capability inventory: Generated 'Playbooks' are intended to be re-executed or followed by the agent in future tasks, potentially leading to the execution of injected instructions.
- Sanitization: There is no evidence of sanitization, filtering, or validation of the workflow content before it is synthesized into a new Playbook.
Audit Metadata