The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a complex multi-agent workflow that inherently involves processing data generated by sub-agents.\n
Ingestion points: The orchestrator agent ingests output and reports from specialist agents like ts-implementer, ts-tester, and code-explorer within the SKILL.md instructions.\n
Boundary markers: The skill explicitly instructs the user to define "The boundary" in every sub-agent prompt to limit their scope and prevent unintended file modifications.\n
Capability inventory: The orchestrator uses the platform's Agent spawning and SendMessage capabilities to execute its plan and aggregate results.\n
Sanitization: The instructions mandate a defensive approach, requiring the orchestrator to "read each report skeptically," verify that success claims were actually tested, and manually triage reviewer findings to filter out noise or conflicting instructions.

orchestrating-ts-agents