ak-worker
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process task instructions from an external 'Agent Kanban' system via the
realmroot toolbox getcommand. If the remote task content is malicious, it could influence the agent's behavior during execution. - Ingestion points: The agent reads task data from
agent-kanban/tasks/<task-id>as described inSKILL.md. - Boundary markers: The skill does not define specific delimiters or instructions to ignore potential commands within the fetched task data.
- Capability inventory: The agent has access to the
realmrootCLI, which allows for modifying task states, adding notes, and executing authenticated GitHub operations viarealmroot exec github. - Sanitization: No sanitization or validation of the fetched task content is specified.
- [COMMAND_EXECUTION]: The skill provides instructions for executing various shell commands using the
realmrootCLI to manage tasks and interact with GitHub. These are standard operations for the skill's intended purpose.
Audit Metadata