pump-ai-agents
Warn
Audited by Gen Agent Trust Hub on Jul 9, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The MCP server configuration instructs agents to execute the
@pump-fun/mcp-serverpackage usingnpx -y. This results in the automatic download and execution of code from the NPM registry at runtime. - [COMMAND_EXECUTION]: The skill defines strict rules for terminal management, requiring agents to execute commands in background terminals and ensure sessions are terminated upon completion.
- [EXTERNAL_DOWNLOADS]: The integration setup triggers external downloads of the
@pump-fun/mcp-serverpackage during the initialization of the MCP server. - [PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection due to its ingestion of multiple external data sources.
- Ingestion points: The agent is instructed to process discovery files (
.well-known/skills.json) and context documents (llms-full.txt) referenced in the skill (SKILL.md). - Boundary markers: No explicit delimiters or instructions to ignore embedded directives are provided for the processed context files.
- Capability inventory: The skill possesses shell command execution capabilities and access to an MCP server (SKILL.md).
- Sanitization: There is no evidence of sanitization or schema validation for the contents of the discovery files before they are incorporated into the agent's context.
Audit Metadata