od-prototype-social-x-post-card
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied input via the $ARGUMENTS placeholder to construct a design brief. This brief is sent to an external tool which returns instructions that the agent is then told to follow for file creation. This creates an ingestion path for untrusted data that influences the agent's file-writing behavior.\n
- Ingestion points: User input enters the context through the $ARGUMENTS variable in SKILL.md.\n
- Boundary markers: There are no delimiters or boundary markers (like 'ignore embedded instructions') surrounding the interpolated user input.\n
- Capability inventory: The skill utilizes tool-based instructions to guide the agent's internal file-editing tools to create or modify local files.\n
- Sanitization: No sanitization, escaping, or validation of the user-provided arguments is performed before they are processed.
Audit Metadata