The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via the <requirement> argument and generated task descriptions. These inputs are interpolated into prompts for delegated agents and verification tasks without sanitization.
Evidence: In A_PLAN and A_EXECUTE, variables like ${requirement}, ${title}, and ${description} are embedded into instructions for the maestro delegate tool.
Evidence: User-derived criteria are used as instructions for agents spawned via spawn_agents_on_csv in the A_VERIFY phase.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform complex environment configuration and task execution.
Evidence: The A_EXECUTE phase dynamically constructs shell commands using maestro delegate to implement code and run verification gates.
Evidence: The skill executes system-level checks such as maestro delegate-config show --json and maestro search to manage its internal state and dependencies.

odyssey-planex