The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided marketing briefs, problem statements, and evidence which act as ingestion points for external data. It implements a strong evidence chain to mitigate risks, including boundary markers in the invariants section of the skill specification and mandatory sanitization through a classification policy (labeling claims as fact, inference, assumption, or unknown). The capability inventory includes reasoning over supplied context and reading packaged references.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute specific local Python scripts, such as ../../scripts/dynamic_router.py and ../../scripts/neural_router.py, to manage task routing and build dependency graphs. These references are part of the intended architecture for coordinating between focused specialists and do not represent arbitrary or untrusted command execution.

marketing-council