higress-wasm-go-plugin

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • Unverifiable Dependencies (MEDIUM): The skill requires downloading several Go packages from 'github.com/higress-group' and 'github.com/tidwall'.
  • Evidence: go get github.com/higress-group/proxy-wasm-go-sdk@go-1.24, go get github.com/higress-group/wasm-go@main, and go get github.com/tidwall/gjson in SKILL.md. These sources are not within the defined trusted organization list.
  • Command Execution & Environment Modification (MEDIUM): The skill modifies the system's Go environment to use a specific regional proxy.
  • Evidence: go env -w GOPROXY=https://proxy.golang.com.cn,direct in SKILL.md. While common for regional optimization, it directs the toolchain to an external third-party proxy for all future package downloads.
  • Indirect Prompt Injection Surface (HIGH): The skill creates components meant to process untrusted external data with high-impact capabilities.
  • Ingestion Points: HTTP request headers (onHttpRequestHeaders), request bodies (ProcessRequestBody), and response bodies (ProcessResponseBody).
  • Boundary Markers: Absent. There are no instructions or templates for delimiting untrusted input or warning the agent/plugin to ignore embedded instructions.
  • Capability Inventory: Capabilities include traffic modification (proxywasm.AddHttpRequestHeader), direct response generation (proxywasm.SendHttpResponse), and external service interactions (wrapper.NewClusterClient, wrapper.NewRedisClusterClient).
  • Sanitization: Absent. The provided templates do not include sanitization, validation, or escaping of data extracted from HTTP streams before using it in logic or downstream calls.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 09:42 AM