The Agent Skills Directory

[SAFE]: The skill defines a core workflow for interacting with a hosted Model Context Protocol (MCP) server. It emphasizes security through its 'Safety boundaries' section, instructing the agent to validate mutations via notis_validate_calls and to treat identifiers as opaque tokens.
[SAFE]: The skill provides clear guidance on preventing credential leakage, explicitly instructing the agent not to place secrets in search queries or tool arguments unless explicitly required by the target tool.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Notis resources (e.g., notis://docs/* and notis_search_tools results) which are used to construct tool calls. While this is an attack surface, the skill includes built-in mitigations by requiring pre-execution validation and restricting local system access.
Ingestion points: Discovery results from notis_search_tools and documentation fetched via notis_fetch or MCP resource URIs.
Boundary markers: The skill defines a 'Safety boundaries' section with explicit rules for the agent to follow when processing these inputs.
Capability inventory: Includes the ability to perform data mutations via notis_execute_write and fetch external content via notis_fetch.
Sanitization: Relies on a server-side validation step (notis_validate_calls) to ensure arguments and tool names are legitimate before execution.

notis-mcp