A2-CAP: A Causal Attribute–Affordance Graph Harness for Frozen VLA Manipulation

Long-horizon robotic manipulation requires reasoning about which object properties enable skills and whether their intended effects occur. We introduce **A2-CAP**, an inference-time harness around a frozen vision-language-action (VLA) policy. Its registry of 12 mechanism-level strategies, plus an LLM fallback, defines a causal attribute–affordance graph whose directed edges encode preconditions and effects. **PiCausal** grounds video evidence in this graph; **CAP-Plan** selects skills, verifies their outcomes, and handles failures without updating the policy. We introduce **A2-CausalHOI**, 1,305 hand–object interaction videos spanning 85 object categories and 256 tasks, with registry-derived target graphs. On a 100-clip graph-grounding sample, PiCausal raises edge F1 from $0.24$ to $0.61$ over a matched VLM-only baseline. On 50 scored EPIC-KITCHENS-100 segments, gains are $+0.43$ to $+0.52$ across three backbones; on 14 segments annotated before viewing model drafts, they are $+0.19$ to $+0.36$. With pre-generated graphs on two dual-arm tasks, the frozen $\pi 0.5$ policy under the harness scores $50.0\%$ task SR over eight T1 rounds versus $0.0\%$ over ten rounds for each of two baselines, under a joint execution-and-planning rubric with partial credit. On T2, which reverses the arm roles, it scores $59.1\%$ task SR over 11 rounds without a baseline comparison.