The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest user-supplied task requirements and transform them into a sequence of tool calls involving other agents, skills, and MCP servers.
Ingestion points: The process begins in 'Step 1: Analyze User Request' where requirements and keywords are extracted from the user's prompt.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions to separate user data from the agent's logic during the planning phase.
Capability inventory: The skill possesses the capability to execute other resources via invokeSkill, invokeAgent, and invokeMCPTool during the final execution phase.
Sanitization: The skill implements a critical human-in-the-loop defense in 'Step 5: Request Approval', which uses AskUserQuestion to require explicit user confirmation of the execution plan before any action is taken.
[COMMAND_EXECUTION]: The primary purpose of this skill is to dynamically orchestrate and execute other functional components of the agent platform.
Evidence: 'Step 6: Execute Plan' involves an asynchronous execution loop that triggers tool invocations based on the plan approved by the user.
[EXTERNAL_DOWNLOADS]: The documentation provides manual installation instructions involving external repositories and package managers.
Evidence: The README provides commands for npm install -g claude-superskills and git clone from a placeholder GitHub repository.
Context: These instructions are for user setup and are not executed automatically by the skill during its operation.

agent-skill-orchestrator