skills/matematicsolutions/awesome-matematic-skills-pl/matematic-mcp-fastmcp-instructions-pl/Gen Agent Trust Hub
matematic-mcp-fastmcp-instructions-pl
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references architectural patterns and source code designs from several well-known public GitHub repositories including Dograh-hq, Firecrawl, and Deepseek-AI. These are used as reference models for implementing server features such as procedural orchestration and capability seams.- [DATA_EXFILTRATION]: The skill provides templates for OpenTelemetry instrumentation that prioritize data privacy. It includes a specific implementation for telemetry attribute filtering using an allowlist and explicitly excludes sensitive error messages (exception.message) from egress to prevent accidental exposure of system paths or PII.- [CREDENTIALS_UNSAFE]: The provided authentication templates in examples/auth.py demonstrate how to handle API keys and Bearer tokens from HTTP headers. The implementation correctly delegates the validation process to a dedicated internal authentication handler rather than performing unsafe local checks or hardcoding secrets.- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for prompt injection by allowing the injection of instructions into the agent's system prompt. 1. Ingestion points: Instructions are injected via the FastMCP constructor in examples/server.py. 2. Boundary markers: The skill recommends separating machine-readable codes from human messages in tool responses. 3. Capability inventory: The skill utilizes Read, Write, and Edit tools. 4. Sanitization: It includes logic to filter and sanitize telemetry output before it leaves the local environment.
Audit Metadata