mcp-tester
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data including MCP tool descriptions and response outputs which represents an attack surface for indirect prompt injection if the source server provides malicious content.
- Ingestion points: Ingests tool descriptions during Phase 1 (Discovery) and captures tool output data during Phase 4 (Response Capture) in
SKILL.md. - Boundary markers: The skill does not define specific delimiters or warnings to ignore instructions embedded within the tool data.
- Capability inventory: The skill has the capability to list and execute any configured
mcp__tool within the Claude Code session, which may include file system or network access depending on the user's server configuration. - Sanitization: No explicit sanitization or filtering of external tool content is performed before processing or reporting.
- [SAFE]: The skill implements a robust safety pattern by requiring the agent to ask for explicit user confirmation before executing any 'Mutating Tools' (operations like create, update, or delete), which mitigates accidental or malicious unauthorized state changes.
- [SAFE]: The documentation provides standard installation and debugging instructions referencing well-known services and official protocol packages (NPM, @modelcontextprotocol/inspector) without executing them automatically.
Audit Metadata