The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a design refinement workflow that ingests and processes suggestions provided by an external MCP tool, creating a surface for indirect prompt injection. If the external tool's output is compromised, it could influence the agent's subsequent actions.
Ingestion points: The prompt argument (user-controlled) and the output_components field from the tool's response as described in SKILL.md.
Boundary markers: No delimiters or explicit instructions are provided to help the agent distinguish between its own system instructions and potentially malicious instructions embedded in the external tool's feedback.
Capability inventory: The skill execution environment is configured with access to stitch*:*, Bash, Read, and Write tools via allowed-tools.
Sanitization: There is no documented validation or sanitization of the text contained within output_components before it is presented to the user or used to trigger subsequent tool calls.

stitch-mcp-edit-screens