The Agent Skills Directory

[SAFE]: The skill consists of developer documentation, templates, and example test files intended to guide the creation of secure MCP servers. No malicious code or instructions were detected.
[CREDENTIALS_SAFE]: Code examples in examples/auth.py and SKILL.md demonstrate how to correctly implement multi-channel authentication (X-API-Key and Bearer tokens) by extracting headers at runtime, rather than using hardcoded secrets. It encourages standard secret management practices like using environment variables for the underlying auth handlers.
[DATA_EXPOSURE]: The skill explicitly addresses data safety by providing a pattern for a telemetry allowlist. It recommends filtering attributes before they leave the process and specifically advises against exporting exception.message to prevent leaking sensitive local system information like file paths, prompts, or provider outputs through observability spans.
[COMMAND_EXECUTION]: The documentation includes standard developer commands for environment setup, testing (pytest), and server registration (claude mcp add). These are benign instructional examples for user-initiated workflows.
[INDIRECT_PROMPT_INJECTION]: The skill provides architectural patterns (such as splitting definitions from providers and creating a coverage tool) that mitigate common agent failure modes. The coverage tool allows the agent to programmatically discover the limits of its data source, reducing the risk of the agent taking actions based on incomplete or misunderstood context.

matematic-mcp-fastmcp-instructions-pl