mcp-integration-reference
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data from MCP tools (e.g., Sentry error reports and stack traces), creating an inherent attack surface where an attacker who can trigger application errors could potentially embed malicious instructions in the resulting logs.
- Ingestion points: Data enters the agent's context via the output of MCP tools such as
mcp__sentry-mcp__list_issues,mcp__sentry-mcp__get_issue, andmcp__sentry-mcp__search_events. - Boundary markers: The current template does not provide instructions for the agent to wrap external tool output in specific delimiters or use "ignore embedded instructions" warnings during the analysis phase.
- Capability inventory: The skill possesses file reading capabilities (
Read) and execution rights for MCP-specific tools (mcp__<your-mcp>__*). - Sanitization: No explicit sanitization, filtering, or validation logic is defined to check ingested log data for control characters or hidden prompt injection patterns.
Audit Metadata