agkan-subtask-direct
Pass
Audited by Gen Agent Trust Hub on Apr 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes various shell commands including the 'agkan' CLI for task management, 'git' for version control operations, and language-specific linting tools such as 'npx', 'bundle', and 'ruff'.
- [PROMPT_INJECTION]: A surface for indirect prompt injection exists because the skill ingests and processes data from external task definitions. * Ingestion points: The skill retrieves task metadata (specifically the 'branch' field) and the task body content via 'agkan task get' and 'agkan task meta'. * Boundary markers: None are present; task-sourced data is used to control shell logic (e.g., git checkout) or recorded into files without delimiters or instructions to ignore embedded commands. * Capability inventory: The skill has access to shell execution, file system modification, and network access via git. * Sanitization: There is no evidence of sanitization or validation of the branch metadata or task body content before it is used in shell operations.
Audit Metadata