craft-discovery-synthesis
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill interpolates user-supplied data into its core instructions using the
$ARGUMENTSvariable. An adversarial user could provide input designed to override the agent's persona or synthesis task. - [INDIRECT_PROMPT_INJECTION]: The skill is specifically designed to process external qualitative data such as feedback and support tickets, which are common vectors for indirect prompt injection attacks.
- Ingestion points: User research data enters the skill via the
$ARGUMENTSvariable inSKILL.md. - Boundary markers: Input data is wrapped in XML-style
<context>tags to differentiate it from instructions. - Capability inventory: The skill has no permissions for tool invocation, file system modification, or network access, which restricts the potential impact of an injection.
- Sanitization: There are no explicit instructions for the model to treat the embedded content as data only or to ignore instructions found within the qualitative notes.
Audit Metadata