od-prototype-article-magazine
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input via the
$ARGUMENTSvariable and interpolates it into a 'Brief' string which is then passed to external tools (prepare_open_design_brief) and used to guide the agent's file-authoring behavior. - Ingestion points: Untrusted data enters the agent context via the
$ARGUMENTSvariable inSKILL.md. - Boundary markers: There are no explicit delimiters (e.g., XML tags or backticks) or 'ignore embedded instructions' warnings surrounding the
$ARGUMENTSinterpolation to prevent it from overriding the skill's intended logic. - Capability inventory: The skill possesses the capability to call multiple tools (
prepare_open_design_brief,register_open_design_artifact) and explicitly instructs the agent to use its own file-editing tools to author new files. - Sanitization: No sanitization, escaping, or validation of the user-provided content is performed before it is used to influence the agent's actions.
Audit Metadata