The Agent Skills Directory

[SAFE]: The skill functions as a foundational routing instruction, providing a decision framework for an AI agent to choose between different toolsets based on project complexity. It performs structural checks on the current working directory for configuration folders (like .planning/) to maintain project state.
[NO_CODE]: No executable scripts, binaries, or shell commands are present in the skill. The Graphviz (dot) blocks are used solely for documenting logical flows and decision trees for the agent to follow.
[INDIRECT_PROMPT_INJECTION]: The skill assesses external context, specifically directory contents and user-provided task descriptions, to make routing decisions. Because the skill lacks any dangerous capabilities—such as file modification, network communication, or arbitrary command execution—this logic-based evaluation poses no security risk. Ingestion points: File system presence checks and user task input. Capability inventory: None. Sanitization: Not applicable given the purely instructional nature.

workflow-routing