tech-radar-boot
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to execute bundled helper scripts (
scripts/radar-search.shandscripts/radar-search.ps1) to perform technology lookups and data processing. - [EXTERNAL_DOWNLOADS]: Retrieves a technology radar dataset from the vendor-owned URL
https://tikalk.com/radar.jsonduring execution. - [INDIRECT_PROMPT_INJECTION]: The skill processes external JSON data, extracts text from HTML blocks (specifically the "Why?" opinion field), and injects this content directly into the agent's response context.
- Ingestion points: External data is fetched from
https://tikalk.com/radar.jsonby the radar search scripts. - Boundary markers: The skill lacks explicit delimiters or instructions for the agent to treat the retrieved technology opinions as untrusted data.
- Capability inventory: The skill executes shell scripts that perform network operations and can influence the agent's decision-making by injecting external rationale.
- Sanitization: Scripts perform basic HTML tag stripping and whitespace normalization but do not sanitize for malicious instructions within the retrieved natural language text.
Audit Metadata