fabric-ambient
Pass
Audited by Gen Agent Trust Hub on Jul 25, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill defines a workflow to provision persistent, autonomous 'actors' that incorporate untrusted input into their system instructions.
- Ingestion points: User-provided
<goal>or[focus]strings are interpolated directly into theinstructionsparameter withinSKILL.mdand passed to the setup script inreferences/setup.md. - Boundary markers: No technical delimiters or 'ignore instructions' warnings are applied to the interpolated input, allowing potential instruction hijacking within the spawned actor's context.
- Capability inventory: The setup script in
references/setup.mdgrants these autonomous agents access to file system tools (read,grep,find,ls) and configures them to trigger based on session events. - Sanitization: The
setup.mdscript performs no validation, filtering, or escaping of the instruction content before using it to create or update an agent.
Audit Metadata