The Agent Skills Directory

Prompt Injection (LOW): The skill uses behavioral triggers (e.g., /orchestrate:spawn) that interpret user input to define agent personas and strategies. This instruction-following design is susceptible to manipulation if a user provides specifically crafted text to override the orchestration logic.
Indirect Prompt Injection (LOW): The skill acts as a meta-orchestrator, creating a significant injection surface.
Ingestion points: Files references/spawn.md and references/task.md accept arbitrary user input via [complex-task] and [action] parameters.
Boundary markers: The delegation templates do not show the use of delimiters (like triple quotes or XML tags) to wrap user input, nor do they include instructions for the agent to ignore instructions embedded within the user data.
Capability inventory: The skill coordinates powerful tools including Task (for spawning new agents), Bash (for shell execution), and Read/Write/Edit (for file system modification).
Sanitization: There is no mention of escaping or validating user input before passing it to sub-agents or the Bash shell.
Command Execution (LOW): Documentation in references/spawn.md explicitly encourages the use of the Bash tool for system-level operations. While necessary for the skill's stated purpose of 'Meta-System Orchestration', the use of shell tools alongside un-sanitized user input increases the overall risk profile.

task-orchestration