In the formulation of your hypotheses you had to assume that this is relevant news to the market with an imidiate effect on prices and volatilities, in other words that a rating action leads to a structural break, not the other way round (that a structural break leads to rating action). The problem in testing will probably be a too small sample within this particular dataset - too few rating actions of countries with relevant size to move the price of commodities.
My idea would be to test three time-series pairwise: First test structural breaks in CDS against Rating Actions to get the lag or lead of the time-series. Second, test structural breaks in CDS against structural breaks in Commodities to get the lag or lead of the time series. Then, if for example you get a lag of 3 Months for Rating Actions to CDS, and a lead of 4 Months for CDS against Commodities, you might conclude that Rating Actions lead relative to structural breaks in commodities.
With regard to your concrete predictions: Whats the reasoning for expecting a short-run dip in oil, and a decline in correlation for gold?