The Agent Skills Directory

[SAFE]: The skill implements a best-practice workflow for software engineering, focusing on research and the reuse of existing solutions to avoid redundancy.
[INDIRECT_PROMPT_INJECTION]: The skill includes a research step that ingests untrusted data from external registries and search results.
Ingestion points: External search results from npm, PyPI, and GitHub (SKILL.md).
Boundary markers: None explicitly defined in the prompt template.
Capability inventory: The workflow permits package installation (e.g., npm install) and informed code generation based on external data.
Sanitization: Relies on the agent's native guardrails for processing untrusted web content.
[DATA_EXPOSURE]: The skill instructions mention accessing ~/.claude/settings.json and ~/.claude/skills/. These references are used to identify available Model Context Protocol (MCP) servers and existing skills to avoid duplicate implementations, which is legitimate discovery behavior within the platform environment.

search-first