kami-landing
Pass
Audited by Gen Agent Trust Hub on May 4, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill functions as a static document generator, producing a single-page HTML file based on user-provided brand briefs. It contains no embedded JavaScript or external executable dependencies, significantly reducing the attack surface.\n- [SAFE]: External network operations are limited to preconnecting and fetching fonts from Google Fonts (fonts.googleapis.com and fonts.gstatic.com), which are well-known technology services and treated as safe sources.\n- [SAFE]: The skill requests only the 'file-write' capability, which is necessary for its primary purpose of saving the generated HTML document. No access to sensitive system directories or credentials was observed.\n- [SAFE]: Indirect Prompt Injection Surface Evaluation:\n
- Ingestion points: The skill accepts user-provided content for brand details and chapter text through prompt inputs in SKILL.md.\n
- Boundary markers: While specific delimiters are absent, the instructions constrain the agent to a narrow design spec that precludes script execution or dynamic behavior.\n
- Capability inventory: The capability surface is limited to 'file-write' for artifact generation.\n
- Sanitization: The agent is instructed to produce a zero-JavaScript environment with fixed CSS tokens, which naturally mitigates the execution of malicious code provided in inputs.\n- [SAFE]: No obfuscation, data exfiltration patterns, or suspicious command execution logic was detected across the instructions or example artifacts.
Audit Metadata