optise-helix-race-audit
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a bundled local script
scripts/fetch_page.pyto perform the webpage analysis and extraction required for the audit. - [EXTERNAL_DOWNLOADS]: The skill performs network requests to fetch external HTML content from user-provided URLs. The included fetch script implements sophisticated security measures, including DNS pinning and IP validation, to prevent Server-Side Request Forgery (SSRF) attacks targeting private or internal networks.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it retrieves and processes untrusted external webpage content.
- Ingestion points: External HTML content is fetched via
scripts/fetch_page.pyand processed inSKILL.mdworkflow steps. - Boundary markers: The
references/operating-principles.mdfile contains a mandatory directive to treat all fetched content as data and to never follow instructions found within those sources. - Capability inventory: The skill has the capability to perform network fetches and execute local scripts, but does not have arbitrary file-write or persistent execution capabilities.
- Sanitization: The fetch script parses specific structural elements (headings, titles, tables, and Schema.org markup) rather than providing the agent with an uncurated raw stream of executable content.
Audit Metadata