The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional markdown and natural language. It does not contain any executable scripts, binary files, or command-line operations that pose a threat to the execution environment.- [NO_CODE]: The skill does not include any source code, software dependencies, or instructions to install third-party packages from registries like npm or PyPI.- [INDIRECT_PROMPT_INJECTION]: The skill uses tools such as WebSearch, WebFetch, and mcp__scout__readable_text to process external information. Ingestion points: Untrusted data enters the agent context via web search and text fetching tools defined in allowed-tools. Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard instructions found within external data. Capability inventory: The skill possesses no destructive capabilities; it cannot write files, execute shell commands, or perform unauthorized network requests (it is restricted to the tools listed). Sanitization: There are no instructions for sanitizing or filtering the fetched data. This surface is evaluated as safe because the skill is purely advisory and lacks the necessary tools to act on malicious instructions injected via external data.

rdops-expert