The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to utilize external research tools, including web search and Context7 MCP for documentation lookups, which presents a surface for indirect prompt injection.
Ingestion points: External content fetched via mandatory web searches and documentation retrieved through the Context7 MCP as defined in SKILL.md (Research-First Approach) and references/mcp-and-research.md.
Boundary markers: Absent. The skill lacks instructions for the agent to delimit, validate, or ignore instructions embedded within external search results or documentation.
Capability inventory: The agent is instructed to perform tasks involving system design and technology evaluation across SKILL.md and various files in the references/ directory; the underlying agent platform typically provides tools (shell, file access) that could be targeted by instructions hidden in external data.
Sanitization: Absent. The skill does not describe any sanitization or filtering procedures for data retrieved from external sources before it enters the agent's context.

solution-architect