The Agent Skills Directory

[SAFE]: The skill provides structural instructions for agent task management, including protocols for identifying user intent, selecting appropriate workflows, and maintaining task synchronization.
[COMMAND_EXECUTION]: The skill references the spawn_agent tool for delegating sub-tasks and documents the !command`` syntax for dynamic context injection. These references are part of the agent's standard operational framework and do not contain executable payloads, unauthorized command strings, or evidence of command injection vulnerabilities.
[DATA_EXPOSURE]: Instructions within the skill direct the agent to access specific project documentation and configuration files (e.g., docs/project-config.json, workflows.json) to maintain task context. This access is scoped to project-relevant information and does not attempt to access sensitive system files or harvest credentials.
[PROMPT_INJECTION]: The skill contains instructions to prioritize its defined protocols and ignore platform-specific mode-switch instructions. These instructions serve as compatibility and robustness measures for the agent's logic rather than attempts to bypass safety guardrails or manipulate the underlying behavioral constraints.

start-workflow