WorldThreatModelHarness

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through its core workflows.
  • Ingestion points: Untrusted data enters the agent context via user-provided ideas in Workflows/TestIdea.md and web research results fetched by the 'Research' skill in Workflows/UpdateModels.md.
  • Boundary markers: The workflow instructions lack explicit delimiters or safety instructions to prevent the agent from obeying instructions embedded within the processed research or user input.
  • Capability inventory: The skill possesses capabilities including local filesystem access (~/.opencode/), shell command execution (curl), and the ability to spawn parallel background agents.
  • Sanitization: No sanitization or validation logic is implemented to filter or escape the external content before it is interpolated into agent prompts.
  • [COMMAND_EXECUTION]: All workflows (TestIdea.md, UpdateModels.md, ViewModels.md) execute shell commands using curl to send POST requests to a local notification service at http://localhost:8888/notify. While restricted to localhost and using structured JSON payloads, this constitutes active subprocess execution within the skill's operational flow.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 07:38 AM