jobbank-search
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSOBFUSCATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted job listings from an external portal (jobbank.dk). It significantly mitigates the risk of indirect prompt injection by implementing aggressive sanitization logic in
cli/src/helpers.ts. ThestripInvisibleandvisibleHtmlfunctions remove non-rendering Unicode characters (such as C0/C1 controls, zero-width characters, and Unicode tag characters) and dangerous HTML elements (scripts, styles, hidden tags) before the data is passed to the AI agent. - [OBFUSCATION]: Static analysis flagged invisible Unicode characters in
cli/src/helpers.tsandcli/tests/hidden-text.test.ts. Investigation reveals these are not used for malicious obfuscation but are part of the skill's defensive layer, used to identify and remove steganographic content from third-party listing data. - [EXTERNAL_DOWNLOADS]: The CLI fetches search results and job details from
https://jobbank.dk. All network operations use validated numeric IDs and enforce hostname checks to prevent potential SSRF or unauthorized redirection.
Audit Metadata