The Agent Skills Directory

[NO_CODE]: This skill consists entirely of instructional markdown and text documentation, containing no executable code or scripts.
[INDIRECT_PROMPT_INJECTION]: The skill defines an orchestration framework that ingests external task context, creating a surface for indirect prompt injection. The instructions include the following safety considerations: (1) Ingestion points: Task context and inputs are identified in SKILL.md under 'Establish the work' and in the handoff template in references/brief-template.md. (2) Boundary markers: The skill mandates the use of explicit 'Scope and boundaries' in handoff briefs to limit sub-worker actions. (3) Capability inventory: The skill notes potential use of native subagents, subprocesses, and Git operations, provided they are authorized in the session. (4) Sanitization: The instructions require agents to 'Check important claims against the artifacts or source evidence' and resolve contradictions before reporting results.

orchestrate