The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided marketing data and evidence to generate design recommendations. It implements several security controls to mitigate potential injection risks:
Ingestion points: Processes user objectives, constraints, and primary evidence through its workflow steps in references/skill-spec.json.
Boundary markers: Requires the agent to classify all material claims using specific status labels: fact, inference, assumption, or unknown as per the Evidence discipline in SKILL.md.
Capability inventory: The skill can invoke internal routing scripts (dynamic_router.py, neural_router.py) and specialized agents (measurement-strategist.md, marketing-skeptic.md) as defined in SKILL.md and references/skill-spec.json.
Sanitization: The Evidence discipline and invariants explicitly prohibit the fabrication of research, tool calls, or metrics, and mandate the separation of inference from evidence.

incrementality-design