The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill dynamically adjusts agent behavior constraints and tool availability based on content read from an external configuration file, creating a surface for instructions within that file to influence the agent's operations.\n
Ingestion points: Reads project configuration data from PROJECT_SETTINGS.md.\n
Boundary markers: The instructions include defensive rules stating that settings do not override the author or canon, and that high-risk work always requires manual approval.\n
Capability inventory: Influences the execution of engines like canon-manager and version-control, as well as task routing via novel-orchestrator.\n
Sanitization: No sanitization or validation logic for the values provided in the configuration file is mentioned in the instructions.

project-settings