workflow-testing

Pass

Audited by Gen Agent Trust Hub on Apr 10, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: Forceful Instructional Directives. The skill utilizes authoritative language, including the terms 'IMPORTANT', 'MUST', and 'Do NOT skip', to enforce a specific sequence of actions. This pattern is characteristic of instructions intended to override an agent's standard behavior or safety guidelines to ensure compliance with a rigid workflow.\n- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill ingests the user's prompt and passes it directly to the /workflow-start testing command as context, which provides an entry point for embedded instructions to influence the agent's actions.\n
  • Ingestion points: The user's prompt is processed and passed to a workflow trigger in SKILL.md.\n
  • Boundary markers: Absent. The skill does not implement delimiters or instructions to ignore commands within the interpolated user content.\n
  • Capability inventory: The skill triggers a workflow consisting of the /workflow-start, /test, and /workflow-end commands.\n
  • Sanitization: Absent. There is no evidence of validation or sanitization applied to the user input before it is used by the workflow engine.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 10, 2026, 07:25 AM
Security Audit — agent-trust-hub — workflow-testing