yy-create-skill
Pass
Audited by Gen Agent Trust Hub on May 11, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill facilitates the generation of instructions (SKILL.md files) based on user-provided input, which creates a surface for indirect prompt injection (Category 8). Malicious input provided during the 'intent capture' phase could lead the AI to generate skills containing instructions that bypass safety filters or override agent behavior.
- Ingestion points: User-provided functional requirements and descriptions for new or updated skills (SKILL.md).
- Boundary markers: The generation logic does not specify the use of delimiters or 'ignore embedded instructions' warnings for user-sourced content.
- Capability inventory: The agent uses file-writing capabilities to create or update directories and markdown files within the
.agents/skillspath. - Sanitization: While the skill includes an 'Acceptance Checklist', it lacks explicit instructions to sanitize or validate user input for prompt injection or system prompt extraction patterns.
Audit Metadata