The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from local project files to determine engine detection and task classification, creating a surface for indirect prompt injection.\n
Ingestion points: The agent is instructed to scan various project files including project.godot, *.uproject, package.json, Cargo.toml, conf.lua, and *.rbxl(x) to identify the engine and version.\n
Boundary markers: There are no specific instructions or delimiters provided to the agent to treat the content of these files as untrusted or to ignore any embedded natural language instructions within them.\n
Capability inventory: While the router skill itself is limited to classification, it dispatches the agent to other specialized skills which likely possess broader file system, network, or execution capabilities.\n
Sanitization: The skill lacks sanitization or validation logic for the content read from these project metadata files.\n- [COMMAND_EXECUTION]: The skill instructs the agent to use system-level CLI tools for environment detection and publishing workflows.\n
Evidence: The agent is guided to run godot --version for version detection in references/engine-detection.md and use butler push for publishing tasks in references/routing-table.md.

router