The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is configured to execute local Python scripts for dynamic routing and neural graph management during task execution.
Evidence: Mentions of ../../scripts/dynamic_router.py, ../../scripts/neural_router.py, and ../../scripts/skill_router.py within SKILL.md and the references/skill-spec.json configuration.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external inputs such as market research, competitive pricing data, and willingness-to-pay signals which could potentially influence the parameters passed to its script-based routing mechanisms.
Ingestion points: Market claims, customer language, and pricing evidence processed in SKILL.md and references/decision-model.md.
Boundary markers: Employs an "Evidence discipline" framework and "invariants" section to classify data as fact, inference, or assumption, providing a layer of defense against malicious data influence.
Capability inventory: Shell execution capabilities for local Python scripts via relative paths.
Sanitization: Requires strict adherence to the local references/skill-spec.json evidence policy to reject neural recommendations that lack support in primary evidence.

pricing-strategy