The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external marketing data and user-supplied evidence. It proactively mitigates injection risks through a strict 'Evidence discipline' and 'Evidence policy' (defined in SKILL.md and references/skill-spec.json) that requires the agent to classify all material claims as fact, inference, assumption, or unknown. It also uses an 'evidence-gate' hook to validate inputs before processing.
[COMMAND_EXECUTION]: Instructions in SKILL.md direct the agent to use local scripts such as ../../scripts/dynamic_router.py and ../../scripts/neural_router.py for routing cross-functional requests and building execution graphs. These are internal architectural components of the agent framework and do not involve untrusted remote execution.
[DYNAMIC_EXECUTION]: The skill utilizes a neural graph (../../neural/graph.json) and specialist agents (../../agents/channel-strategist.md) to handle specific sub-tasks. The selection and invocation of these components are governed by the skill's internal logic and completion gates, maintaining a controlled execution environment.

media-strategy