next-task
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill ingests untrusted data from project plan files located in
docs/plans/and interpolates task descriptions directly into agent instructions. This creates a surface for indirect prompt injection where malicious content in a plan file could subvert agent behavior. 1. Ingestion point: Reads markdown files from thedocs/plans/directory. 2. Boundary markers: Absent; task content is directly passed to the agent prompt. 3. Capability inventory: Executes shell commands (make test), creates agent teams (TeamCreate), and manages tasks. 4. Sanitization: Absent. - [COMMAND_EXECUTION]: The skill utilizes the
!command`` syntax (dynamic context injection) to execute shell-based file discovery (find) at load time to populate context. It also specifies the execution ofmake testfor task verification, which runs project-specific logic.
Audit Metadata