social-card-gen
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external URLs or local files to generate social media posts. This creates a vulnerability surface where malicious instructions embedded in the input could influence an AI agent that subsequently handles the output.
- Ingestion points: The
generate.jsscript reads input via--file,--url, and--textcommand-line arguments. - Boundary markers: No boundary markers or clear instructions are used to separate input content from processing logic, potentially allowing embedded instructions to persist in the generated output.
- Capability inventory: The skill possesses file-reading (
fs.readFileSync), file-writing (fs.writeFileSync), and network-fetching (fetch) capabilities. - Sanitization: Input text undergoes basic sanitization via
stripMarkdownAndHtml, which removes Markdown and HTML syntax but does not prevent semantic instruction injection. - [DATA_EXFILTRATION]: The
generate.jsscript utilizes thefetchAPI to retrieve data from arbitrary, user-supplied URLs. While this is a documented feature for content ingestion, it grants the skill the ability to perform outbound network operations to non-whitelisted domains.
Audit Metadata