global-chat-agent-discovery
Warn
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to query and potentially install the '@global-chat/mcp-server' package from the npm registry, which is a third-party community resource.
- [REMOTE_CODE_EXECUTION]: The skill suggests a configuration pattern to execute the external package via 'npx'. While it recommends using a 'REVIEWED_VERSION', this pattern facilitates the execution of remote code from an unverified community source.
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for ingesting data from external registries and a directory service ('global-chat.io'). * Ingestion points: Search results and server metadata retrieved via the MCP server or web directory. * Boundary markers: Not specified. * Capability inventory: Potential installation and execution of discovered servers. * Sanitization: None mentioned for processing retrieved descriptions or manifests.
- [METADATA_POISONING]: The skill documentation explicitly states that the project's public source repository returned a 404 error as of 2026-10-14. This discrepancy between the available package and the missing source code poses a significant supply-chain risk.
Audit Metadata