The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from the external codebase (e.g., db/schema.rb, config/routes.rb, and various implementation files). This creates a surface for indirect prompt injection if the project files contain malicious instructions intended to manipulate the agent's behavior.
Ingestion points: Project files including schema, routes, and neighboring class implementations (controllers, models, etc.) are loaded into the agent's context.
Boundary markers: The skill enforces a structured 'Context Summary' output and requires citations (path and line numbers) for all files read, providing transparency for the user to verify sources.
Capability inventory: The skill utilizes file reading and grep tools to inspect the local filesystem, as well as a get_project_context tool.
Sanitization: The instructions do not specify any sanitization or validation of the content read from the files before it is processed or summarized.

load-context