using-skills
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill uses authoritative directives such as 'Iron Law', 'mandatory', and 'MUST follow' to strictly override the agent's standard reasoning process. It forces the agent to prioritize skill invocation over its default response behavior for nearly any task, effectively constraining its operational autonomy.
- [INDIRECT_PROMPT_INJECTION]: The skill defines a mandatory evaluation process for user-provided tasks that lacks proper security controls, creating an attack surface where embedded instructions could influence skill selection.
- Ingestion points: The workflow is triggered by any 'Task Received' and involves scanning the content against a skill taxonomy (SKILL.md).
- Boundary markers: There are no instructions to use delimiters or ignore instructions within the task content during the applicability assessment.
- Capability inventory: The skill triggers the invocation of other skills, which potentially grant access to broader system tools and execution capabilities.
- Sanitization: The process does not include any validation or filtering of user-provided task content before it is processed by the agent's logic.
Audit Metadata